Merge tag 'for-linus-v3.8-rc1' of git://oss.sgi.com/xfs/xfs

Pull xfs update from Ben Myers:
 "There is plenty going on, including the cleanup of xfssyncd, metadata
  verifiers, CRC infrastructure for the log, tracking of inodes with
  speculative allocation, a cleanup of xfs_fs_subr.c, fixes for
  XFS_IOC_ZERO_RANGE, and important fix related to log replay (only
  update the last_sync_lsn when a transaction completes), a fix for
  deadlock on AGF buffers, documentation and comment updates, and a few
  more cleanups and fixes.

  Details:
   - remove the xfssyncd mess
   - only update the last_sync_lsn when a transaction completes
   - zero allocation_args on the kernel stack
   - fix AGF/alloc workqueue deadlock
   - silence uninitialised f.file warning
   - Update inode alloc comments
   - Update mount options documentation
   - report projid32bit feature in geometry call
   - speculative preallocation inode tracking
   - fix attr tree double split corruption
   - fix broken error handling in xfs_vm_writepage
   - drop buffer io reference when a bad bio is built
   - add more attribute tree trace points
   - growfs infrastructure changes for 3.8
   - fs/xfs/xfs_fs_subr.c die die die
   - add CRC infrastructure
   - add CRC checks to the log
   - Remove description of nodelaylog mount option from xfs.txt
   - inode allocation should use unmapped buffers
   - byte range granularity for XFS_IOC_ZERO_RANGE
   - fix direct IO nested transaction deadlock
   - fix stray dquot unlock when reclaiming dquots
   - fix sparse reported log CRC endian issue"

Fix up trivial conflict in fs/xfs/xfs_fsops.c due to the same patch
having been applied twice (commits eaef854335ce and 1375cb65e87b: "xfs:
growfs: don't read garbage for new secondary superblocks") with later
updates to the affected code in the XFS tree.

* tag 'for-linus-v3.8-rc1' of git://oss.sgi.com/xfs/xfs: (78 commits)
  xfs: fix sparse reported log CRC endian issue
  xfs: fix stray dquot unlock when reclaiming dquots
  xfs: fix direct IO nested transaction deadlock.
  xfs: byte range granularity for XFS_IOC_ZERO_RANGE
  xfs: inode allocation should use unmapped buffers.
  xfs: Remove the description of nodelaylog mount option from xfs.txt
  xfs: add CRC checks to the log
  xfs: add CRC infrastructure
  xfs: convert buffer verifiers to an ops structure.
  xfs: connect up write verifiers to new buffers
  xfs: add pre-write metadata buffer verifier callbacks
  xfs: add buffer pre-write callback
  xfs: Add verifiers to dir2 data readahead.
  xfs: add xfs_da_node verification
  xfs: factor and verify attr leaf reads
  xfs: factor dir2 leaf read
  xfs: factor out dir2 data block reading
  xfs: factor dir2 free block reading
  xfs: verify dir2 block format buffers
  xfs: factor dir2 block read operations
  ...

Documentation/filesystems/xfs.txt

@@ -43,7 +43,7 @@ When mounting an XFS filesystem, the following options are accepted.
 	Issue command to let the block device reclaim space freed by the
 	filesystem.  This is useful for SSD devices, thinly provisioned
 	LUNs and virtual machine images, but may have a performance
-	impact.  This option is incompatible with the nodelaylog option.
+	impact.
 
   dmapi
 	Enable the DMAPI (Data Management API) event callouts.
@@ -72,8 +72,15 @@ When mounting an XFS filesystem, the following options are accepted.
 	Indicates that XFS is allowed to create inodes at any location
 	in the filesystem, including those which will result in inode
 	numbers occupying more than 32 bits of significance.  This is
-	provided for backwards compatibility, but causes problems for
-	backup applications that cannot handle large inode numbers.
+	the default allocation option. Applications which do not handle
+	inode numbers bigger than 32 bits, should use inode32 option.
+
+  inode32
+	Indicates that XFS is limited to create inodes at locations which
+	will not result in inode numbers with more than 32 bits of
+	significance. This is provided for backwards compatibility, since
+	64 bits inode numbers might cause problems for some applications
+	that cannot handle large inode numbers.
 
   largeio/nolargeio
 	If "nolargeio" is specified, the optimal I/O reported in

fs/xfs/Kconfig

@@ -2,6 +2,7 @@ config XFS_FS
 	tristate "XFS filesystem support"
 	depends on BLOCK
 	select EXPORTFS
+	select LIBCRC32C
 	help
 	  XFS is a high performance journaling filesystem which originated
 	  on the SGI IRIX platform.  It is completely multi-threaded, can

fs/xfs/Makefile

@@ -37,9 +37,8 @@ xfs-y				+= xfs_aops.o \
 				   xfs_file.o \
 				   xfs_filestream.o \
 				   xfs_fsops.o \
-				   xfs_fs_subr.o \
 				   xfs_globals.o \
-				   xfs_iget.o \
+				   xfs_icache.o \
 				   xfs_ioctl.o \
 				   xfs_iomap.o \
 				   xfs_iops.o \
@@ -47,7 +46,6 @@ xfs-y				+= xfs_aops.o \
 				   xfs_message.o \
 				   xfs_mru_cache.o \
 				   xfs_super.o \
-				   xfs_sync.o \
 				   xfs_xattr.o \
 				   xfs_rename.o \
 				   xfs_utils.o \

fs/xfs/uuid.h

@@ -26,4 +26,10 @@ extern int uuid_is_nil(uuid_t *uuid);
 extern int uuid_equal(uuid_t *uuid1, uuid_t *uuid2);
 extern void uuid_getnodeuniq(uuid_t *uuid, int fsid [2]);
 
+static inline void
+uuid_copy(uuid_t *dst, uuid_t *src)
+{
+	memcpy(dst, src, sizeof(uuid_t));
+}
+
 #endif	/* __XFS_SUPPORT_UUID_H__ */

fs/xfs/xfs_ag.h

@@ -108,6 +108,8 @@ typedef struct xfs_agf {
 extern int xfs_read_agf(struct xfs_mount *mp, struct xfs_trans *tp,
 			xfs_agnumber_t agno, int flags, struct xfs_buf **bpp);
 
+extern const struct xfs_buf_ops xfs_agf_buf_ops;
+
 /*
  * Size of the unlinked inode hash table in the agi.
  */
@@ -161,6 +163,8 @@ typedef struct xfs_agi {
 extern int xfs_read_agi(struct xfs_mount *mp, struct xfs_trans *tp,
 				xfs_agnumber_t agno, struct xfs_buf **bpp);
 
+extern const struct xfs_buf_ops xfs_agi_buf_ops;
+
 /*
  * The third a.g. block contains the a.g. freelist, an array
  * of block pointers to blocks owned by the allocation btree code.
@@ -233,6 +237,7 @@ typedef struct xfs_perag {
 #define XFS_ICI_NO_TAG		(-1)	/* special flag for an untagged lookup
 					   in xfs_inode_ag_iterator */
 #define XFS_ICI_RECLAIM_TAG	0	/* inode is to be reclaimed */
+#define XFS_ICI_EOFBLOCKS_TAG	1	/* inode has blocks beyond EOF */
 
 #define	XFS_AG_MAXLEVELS(mp)		((mp)->m_ag_maxlevels)
 #define	XFS_MIN_FREELIST_RAW(bl,cl,mp)	\

fs/xfs/xfs_alloc.c

@@ -430,6 +430,60 @@ xfs_alloc_fixup_trees(
 	return 0;
 }
 
+static void
+xfs_agfl_verify(
+	struct xfs_buf	*bp)
+{
+#ifdef WHEN_CRCS_COME_ALONG
+	/*
+	 * we cannot actually do any verification of the AGFL because mkfs does
+	 * not initialise the AGFL to zero or NULL. Hence the only valid part of
+	 * the AGFL is what the AGF says is active. We can't get to the AGF, so
+	 * we can't verify just those entries are valid.
+	 *
+	 * This problem goes away when the CRC format change comes along as that
+	 * requires the AGFL to be initialised by mkfs. At that point, we can
+	 * verify the blocks in the agfl -active or not- lie within the bounds
+	 * of the AG. Until then, just leave this check ifdef'd out.
+	 */
+	struct xfs_mount *mp = bp->b_target->bt_mount;
+	struct xfs_agfl	*agfl = XFS_BUF_TO_AGFL(bp);
+	int		agfl_ok = 1;
+
+	int		i;
+
+	for (i = 0; i < XFS_AGFL_SIZE(mp); i++) {
+		if (be32_to_cpu(agfl->agfl_bno[i]) == NULLAGBLOCK ||
+		    be32_to_cpu(agfl->agfl_bno[i]) >= mp->m_sb.sb_agblocks)
+			agfl_ok = 0;
+	}
+
+	if (!agfl_ok) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, agfl);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+#endif
+}
+
+static void
+xfs_agfl_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_agfl_verify(bp);
+}
+
+static void
+xfs_agfl_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_agfl_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_agfl_buf_ops = {
+	.verify_read = xfs_agfl_read_verify,
+	.verify_write = xfs_agfl_write_verify,
+};
+
 /*
  * Read in the allocation group free block array.
  */
@@ -447,7 +501,7 @@ xfs_alloc_read_agfl(
 	error = xfs_trans_read_buf(
 			mp, tp, mp->m_ddev_targp,
 			XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
-			XFS_FSS_TO_BB(mp, 1), 0, &bp);
+			XFS_FSS_TO_BB(mp, 1), 0, &bp, &xfs_agfl_buf_ops);
 	if (error)
 		return error;
 	ASSERT(!xfs_buf_geterror(bp));
@@ -2091,6 +2145,63 @@ xfs_alloc_put_freelist(
 	return 0;
 }
 
+static void
+xfs_agf_verify(
+	struct xfs_buf	*bp)
+ {
+	struct xfs_mount *mp = bp->b_target->bt_mount;
+	struct xfs_agf	*agf;
+	int		agf_ok;
+
+	agf = XFS_BUF_TO_AGF(bp);
+
+	agf_ok = agf->agf_magicnum == cpu_to_be32(XFS_AGF_MAGIC) &&
+		XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
+		be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
+		be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
+		be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
+		be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp);
+
+	/*
+	 * during growfs operations, the perag is not fully initialised,
+	 * so we can't use it for any useful checking. growfs ensures we can't
+	 * use it by using uncached buffers that don't have the perag attached
+	 * so we can detect and avoid this problem.
+	 */
+	if (bp->b_pag)
+		agf_ok = agf_ok && be32_to_cpu(agf->agf_seqno) ==
+						bp->b_pag->pag_agno;
+
+	if (xfs_sb_version_haslazysbcount(&mp->m_sb))
+		agf_ok = agf_ok && be32_to_cpu(agf->agf_btreeblks) <=
+						be32_to_cpu(agf->agf_length);
+
+	if (unlikely(XFS_TEST_ERROR(!agf_ok, mp, XFS_ERRTAG_ALLOC_READ_AGF,
+			XFS_RANDOM_ALLOC_READ_AGF))) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, agf);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_agf_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_agf_verify(bp);
+}
+
+static void
+xfs_agf_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_agf_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_agf_buf_ops = {
+	.verify_read = xfs_agf_read_verify,
+	.verify_write = xfs_agf_write_verify,
+};
+
 /*
  * Read in the allocation group header (free/alloc section).
  */
@@ -2102,44 +2213,19 @@ xfs_read_agf(
 	int			flags,	/* XFS_BUF_ */
 	struct xfs_buf		**bpp)	/* buffer for the ag freelist header */
 {
-	struct xfs_agf	*agf;		/* ag freelist header */
-	int		agf_ok;		/* set if agf is consistent */
 	int		error;
 
 	ASSERT(agno != NULLAGNUMBER);
 	error = xfs_trans_read_buf(
 			mp, tp, mp->m_ddev_targp,
 			XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
-			XFS_FSS_TO_BB(mp, 1), flags, bpp);
+			XFS_FSS_TO_BB(mp, 1), flags, bpp, &xfs_agf_buf_ops);
 	if (error)
 		return error;
 	if (!*bpp)
 		return 0;
 
 	ASSERT(!(*bpp)->b_error);
-	agf = XFS_BUF_TO_AGF(*bpp);
-
-	/*
-	 * Validate the magic number of the agf block.
-	 */
-	agf_ok =
-		agf->agf_magicnum == cpu_to_be32(XFS_AGF_MAGIC) &&
-		XFS_AGF_GOOD_VERSION(be32_to_cpu(agf->agf_versionnum)) &&
-		be32_to_cpu(agf->agf_freeblks) <= be32_to_cpu(agf->agf_length) &&
-		be32_to_cpu(agf->agf_flfirst) < XFS_AGFL_SIZE(mp) &&
-		be32_to_cpu(agf->agf_fllast) < XFS_AGFL_SIZE(mp) &&
-		be32_to_cpu(agf->agf_flcount) <= XFS_AGFL_SIZE(mp) &&
-		be32_to_cpu(agf->agf_seqno) == agno;
-	if (xfs_sb_version_haslazysbcount(&mp->m_sb))
-		agf_ok = agf_ok && be32_to_cpu(agf->agf_btreeblks) <=
-						be32_to_cpu(agf->agf_length);
-	if (unlikely(XFS_TEST_ERROR(!agf_ok, mp, XFS_ERRTAG_ALLOC_READ_AGF,
-			XFS_RANDOM_ALLOC_READ_AGF))) {
-		XFS_CORRUPTION_ERROR("xfs_alloc_read_agf",
-				     XFS_ERRLEVEL_LOW, mp, agf);
-		xfs_trans_brelse(tp, *bpp);
-		return XFS_ERROR(EFSCORRUPTED);
-	}
 	xfs_buf_set_ref(*bpp, XFS_AGF_REF);
 	return 0;
 }

fs/xfs/xfs_alloc.h

@@ -231,4 +231,7 @@ xfs_alloc_get_rec(
 	xfs_extlen_t		*len,	/* output: length of extent */
 	int			*stat);	/* output: success/failure */
 
+extern const struct xfs_buf_ops xfs_agf_buf_ops;
+extern const struct xfs_buf_ops xfs_agfl_buf_ops;
+
 #endif	/* __XFS_ALLOC_H__ */

fs/xfs/xfs_alloc_btree.c

@@ -272,6 +272,82 @@ xfs_allocbt_key_diff(
 	return (__int64_t)be32_to_cpu(kp->ar_startblock) - rec->ar_startblock;
 }
 
+static void
+xfs_allocbt_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
+	struct xfs_perag	*pag = bp->b_pag;
+	unsigned int		level;
+	int			sblock_ok; /* block passes checks */
+
+	/*
+	 * magic number and level verification
+	 *
+	 * During growfs operations, we can't verify the exact level as the
+	 * perag is not fully initialised and hence not attached to the buffer.
+	 * In this case, check against the maximum tree depth.
+	 */
+	level = be16_to_cpu(block->bb_level);
+	switch (block->bb_magic) {
+	case cpu_to_be32(XFS_ABTB_MAGIC):
+		if (pag)
+			sblock_ok = level < pag->pagf_levels[XFS_BTNUM_BNOi];
+		else
+			sblock_ok = level < mp->m_ag_maxlevels;
+		break;
+	case cpu_to_be32(XFS_ABTC_MAGIC):
+		if (pag)
+			sblock_ok = level < pag->pagf_levels[XFS_BTNUM_CNTi];
+		else
+			sblock_ok = level < mp->m_ag_maxlevels;
+		break;
+	default:
+		sblock_ok = 0;
+		break;
+	}
+
+	/* numrecs verification */
+	sblock_ok = sblock_ok &&
+		be16_to_cpu(block->bb_numrecs) <= mp->m_alloc_mxr[level != 0];
+
+	/* sibling pointer verification */
+	sblock_ok = sblock_ok &&
+		(block->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK) ||
+		 be32_to_cpu(block->bb_u.s.bb_leftsib) < mp->m_sb.sb_agblocks) &&
+		block->bb_u.s.bb_leftsib &&
+		(block->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK) ||
+		 be32_to_cpu(block->bb_u.s.bb_rightsib) < mp->m_sb.sb_agblocks) &&
+		block->bb_u.s.bb_rightsib;
+
+	if (!sblock_ok) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_allocbt_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_allocbt_verify(bp);
+}
+
+static void
+xfs_allocbt_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_allocbt_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_allocbt_buf_ops = {
+	.verify_read = xfs_allocbt_read_verify,
+	.verify_write = xfs_allocbt_write_verify,
+};
+
+
 #ifdef DEBUG
 STATIC int
 xfs_allocbt_keys_inorder(
@@ -327,6 +403,7 @@ static const struct xfs_btree_ops xfs_allocbt_ops = {
 	.init_rec_from_cur	= xfs_allocbt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_allocbt_init_ptr_from_cur,
 	.key_diff		= xfs_allocbt_key_diff,
+	.buf_ops		= &xfs_allocbt_buf_ops,
 #ifdef DEBUG
 	.keys_inorder		= xfs_allocbt_keys_inorder,
 	.recs_inorder		= xfs_allocbt_recs_inorder,

fs/xfs/xfs_alloc_btree.h

@@ -93,4 +93,6 @@ extern struct xfs_btree_cur *xfs_allocbt_init_cursor(struct xfs_mount *,
 		xfs_agnumber_t, xfs_btnum_t);
 extern int xfs_allocbt_maxrecs(struct xfs_mount *, int, int);
 
+extern const struct xfs_buf_ops xfs_allocbt_buf_ops;
+
 #endif	/* __XFS_ALLOC_BTREE_H__ */

fs/xfs/xfs_aops.c

@@ -124,7 +124,7 @@ xfs_setfilesize_trans_alloc(
 	ioend->io_append_trans = tp;
 
 	/*
-	 * We will pass freeze protection with a transaction.  So tell lockdep
+	 * We may pass freeze protection with a transaction.  So tell lockdep
 	 * we released it.
 	 */
 	rwsem_release(&ioend->io_inode->i_sb->s_writers.lock_map[SB_FREEZE_FS-1],
@@ -149,11 +149,13 @@ xfs_setfilesize(
 	xfs_fsize_t		isize;
 
 	/*
-	 * The transaction was allocated in the I/O submission thread,
-	 * thus we need to mark ourselves as beeing in a transaction
-	 * manually.
+	 * The transaction may have been allocated in the I/O submission thread,
+	 * thus we need to mark ourselves as beeing in a transaction manually.
+	 * Similarly for freeze protection.
 	 */
 	current_set_flags_nested(&tp->t_pflags, PF_FSTRANS);
+	rwsem_acquire_read(&VFS_I(ip)->i_sb->s_writers.lock_map[SB_FREEZE_FS-1],
+			   0, 1, _THIS_IP_);
 
 	xfs_ilock(ip, XFS_ILOCK_EXCL);
 	isize = xfs_new_eof(ip, ioend->io_offset + ioend->io_size);
@@ -187,7 +189,8 @@ xfs_finish_ioend(
 
 		if (ioend->io_type == XFS_IO_UNWRITTEN)
 			queue_work(mp->m_unwritten_workqueue, &ioend->io_work);
-		else if (ioend->io_append_trans)
+		else if (ioend->io_append_trans ||
+			 (ioend->io_isdirect && xfs_ioend_is_append(ioend)))
 			queue_work(mp->m_data_workqueue, &ioend->io_work);
 		else
 			xfs_destroy_ioend(ioend);
@@ -205,15 +208,6 @@ xfs_end_io(
 	struct xfs_inode *ip = XFS_I(ioend->io_inode);
 	int		error = 0;
 
-	if (ioend->io_append_trans) {
-		/*
-		 * We've got freeze protection passed with the transaction.
-		 * Tell lockdep about it.
-		 */
-		rwsem_acquire_read(
-			&ioend->io_inode->i_sb->s_writers.lock_map[SB_FREEZE_FS-1],
-			0, 1, _THIS_IP_);
-	}
 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
 		ioend->io_error = -EIO;
 		goto done;
@@ -226,35 +220,31 @@ xfs_end_io(
 	 * range to normal written extens after the data I/O has finished.
 	 */
 	if (ioend->io_type == XFS_IO_UNWRITTEN) {
+		error = xfs_iomap_write_unwritten(ip, ioend->io_offset,
+						  ioend->io_size);
+	} else if (ioend->io_isdirect && xfs_ioend_is_append(ioend)) {
 		/*
-		 * For buffered I/O we never preallocate a transaction when
-		 * doing the unwritten extent conversion, but for direct I/O
-		 * we do not know if we are converting an unwritten extent
-		 * or not at the point where we preallocate the transaction.
+		 * For direct I/O we do not know if we need to allocate blocks
+		 * or not so we can't preallocate an append transaction as that
+		 * results in nested reservations and log space deadlocks. Hence
+		 * allocate the transaction here. While this is sub-optimal and
+		 * can block IO completion for some time, we're stuck with doing
+		 * it this way until we can pass the ioend to the direct IO
+		 * allocation callbacks and avoid nesting that way.
 		 */
-		if (ioend->io_append_trans) {
-			ASSERT(ioend->io_isdirect);
-
-			current_set_flags_nested(
-				&ioend->io_append_trans->t_pflags, PF_FSTRANS);
-			xfs_trans_cancel(ioend->io_append_trans, 0);
-		}
-
-		error = xfs_iomap_write_unwritten(ip, ioend->io_offset,
-						 ioend->io_size);
-		if (error) {
-			ioend->io_error = -error;
+		error = xfs_setfilesize_trans_alloc(ioend);
+		if (error)
 			goto done;
-		}
+		error = xfs_setfilesize(ioend);
 	} else if (ioend->io_append_trans) {
 		error = xfs_setfilesize(ioend);
-		if (error)
-			ioend->io_error = -error;
 	} else {
 		ASSERT(!xfs_ioend_is_append(ioend));
 	}
 
 done:
+	if (error)
+		ioend->io_error = -error;
 	xfs_destroy_ioend(ioend);
 }
 
@@ -1432,25 +1422,21 @@ xfs_vm_direct_IO(
 		size_t size = iov_length(iov, nr_segs);
 
 		/*
-		 * We need to preallocate a transaction for a size update
-		 * here.  In the case that this write both updates the size
-		 * and converts at least on unwritten extent we will cancel
-		 * the still clean transaction after the I/O has finished.
+		 * We cannot preallocate a size update transaction here as we
+		 * don't know whether allocation is necessary or not. Hence we
+		 * can only tell IO completion that one is necessary if we are
+		 * not doing unwritten extent conversion.
 		 */
 		iocb->private = ioend = xfs_alloc_ioend(inode, XFS_IO_DIRECT);
-		if (offset + size > XFS_I(inode)->i_d.di_size) {
-			ret = xfs_setfilesize_trans_alloc(ioend);
-			if (ret)
-				goto out_destroy_ioend;
+		if (offset + size > XFS_I(inode)->i_d.di_size)
 			ioend->io_isdirect = 1;
-		}
 
 		ret = __blockdev_direct_IO(rw, iocb, inode, bdev, iov,
 					    offset, nr_segs,
 					    xfs_get_blocks_direct,
 					    xfs_end_io_direct_write, NULL, 0);
 		if (ret != -EIOCBQUEUED && iocb->private)
-			goto out_trans_cancel;
+			goto out_destroy_ioend;
 	} else {
 		ret = __blockdev_direct_IO(rw, iocb, inode, bdev, iov,
 					    offset, nr_segs,
@@ -1460,15 +1446,6 @@ xfs_vm_direct_IO(
 
 	return ret;
 
-out_trans_cancel:
-	if (ioend->io_append_trans) {
-		current_set_flags_nested(&ioend->io_append_trans->t_pflags,
-					 PF_FSTRANS);
-		rwsem_acquire_read(
-			&inode->i_sb->s_writers.lock_map[SB_FREEZE_FS-1],
-			0, 1, _THIS_IP_);
-		xfs_trans_cancel(ioend->io_append_trans, 0);
-	}
 out_destroy_ioend:
 	xfs_destroy_ioend(ioend);
 	return ret;
@@ -1641,7 +1618,7 @@ xfs_vm_bmap(
 
 	trace_xfs_vm_bmap(XFS_I(inode));
 	xfs_ilock(ip, XFS_IOLOCK_SHARED);
-	xfs_flush_pages(ip, (xfs_off_t)0, -1, 0, FI_REMAPF);
+	filemap_write_and_wait(mapping);
 	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
 	return generic_block_bmap(mapping, block, xfs_get_blocks);
 }

fs/xfs/xfs_attr.c

@@ -903,11 +903,9 @@ xfs_attr_leaf_addname(xfs_da_args_t *args)
 	 */
 	dp = args->dp;
 	args->blkno = 0;
-	error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
-					     XFS_ATTR_FORK);
+	error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
 	if (error)
-		return(error);
-	ASSERT(bp != NULL);
+		return error;
 
 	/*
 	 * Look up the given attribute in the leaf block.  Figure out if
@@ -1031,12 +1029,12 @@ xfs_attr_leaf_addname(xfs_da_args_t *args)
 		 * Read in the block containing the "old" attr, then
 		 * remove the "old" attr from that block (neat, huh!)
 		 */
-		error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1,
-						     &bp, XFS_ATTR_FORK);
+		error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno,
+					   -1, &bp);
 		if (error)
-			return(error);
-		ASSERT(bp != NULL);
-		(void)xfs_attr_leaf_remove(bp, args);
+			return error;
+
+		xfs_attr_leaf_remove(bp, args);
 
 		/*
 		 * If the result is small enough, shrink it all into the inode.
@@ -1100,20 +1098,17 @@ xfs_attr_leaf_removename(xfs_da_args_t *args)
 	 */
 	dp = args->dp;
 	args->blkno = 0;
-	error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
-					     XFS_ATTR_FORK);
-	if (error) {
-		return(error);
-	}
+	error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
+	if (error)
+		return error;
 
-	ASSERT(bp != NULL);
 	error = xfs_attr_leaf_lookup_int(bp, args);
 	if (error == ENOATTR) {
 		xfs_trans_brelse(args->trans, bp);
 		return(error);
 	}
 
-	(void)xfs_attr_leaf_remove(bp, args);
+	xfs_attr_leaf_remove(bp, args);
 
 	/*
 	 * If the result is small enough, shrink it all into the inode.
@@ -1155,12 +1150,12 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
 	struct xfs_buf *bp;
 	int error;
 
+	trace_xfs_attr_leaf_get(args);
+
 	args->blkno = 0;
-	error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
-					     XFS_ATTR_FORK);
+	error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
 	if (error)
-		return(error);
-	ASSERT(bp != NULL);
+		return error;
 
 	error = xfs_attr_leaf_lookup_int(bp, args);
 	if (error != EEXIST)  {
@@ -1181,22 +1176,15 @@ xfs_attr_leaf_get(xfs_da_args_t *args)
 STATIC int
 xfs_attr_leaf_list(xfs_attr_list_context_t *context)
 {
-	xfs_attr_leafblock_t *leaf;
 	int error;
 	struct xfs_buf *bp;
 
+	trace_xfs_attr_leaf_list(context);
+
 	context->cursor->blkno = 0;
-	error = xfs_da_read_buf(NULL, context->dp, 0, -1, &bp, XFS_ATTR_FORK);
+	error = xfs_attr_leaf_read(NULL, context->dp, 0, -1, &bp);
 	if (error)
 		return XFS_ERROR(error);
-	ASSERT(bp != NULL);
-	leaf = bp->b_addr;
-	if (unlikely(leaf->hdr.info.magic != cpu_to_be16(XFS_ATTR_LEAF_MAGIC))) {
-		XFS_CORRUPTION_ERROR("xfs_attr_leaf_list", XFS_ERRLEVEL_LOW,
-				     context->dp->i_mount, leaf);
-		xfs_trans_brelse(NULL, bp);
-		return XFS_ERROR(EFSCORRUPTED);
-	}
 
 	error = xfs_attr_leaf_list_int(bp, context);
 	xfs_trans_brelse(NULL, bp);
@@ -1600,12 +1588,9 @@ xfs_attr_node_removename(xfs_da_args_t *args)
 		ASSERT(state->path.blk[0].bp);
 		state->path.blk[0].bp = NULL;
 
-		error = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp,
-						     XFS_ATTR_FORK);
+		error = xfs_attr_leaf_read(args->trans, args->dp, 0, -1, &bp);
 		if (error)
 			goto out;
-		ASSERT((((xfs_attr_leafblock_t *)bp->b_addr)->hdr.info.magic) ==
-		       cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 
 		if ((forkoff = xfs_attr_shortform_allfit(bp, dp))) {
 			xfs_bmap_init(args->flist, args->firstblock);
@@ -1653,6 +1638,8 @@ xfs_attr_fillstate(xfs_da_state_t *state)
 	xfs_da_state_blk_t *blk;
 	int level;
 
+	trace_xfs_attr_fillstate(state->args);
+
 	/*
 	 * Roll down the "path" in the state structure, storing the on-disk
 	 * block number for those buffers in the "path".
@@ -1699,6 +1686,8 @@ xfs_attr_refillstate(xfs_da_state_t *state)
 	xfs_da_state_blk_t *blk;
 	int level, error;
 
+	trace_xfs_attr_refillstate(state->args);
+
 	/*
 	 * Roll down the "path" in the state structure, storing the on-disk
 	 * block number for those buffers in the "path".
@@ -1707,7 +1696,7 @@ xfs_attr_refillstate(xfs_da_state_t *state)
 	ASSERT((path->active >= 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
 	for (blk = path->blk, level = 0; level < path->active; blk++, level++) {
 		if (blk->disk_blkno) {
-			error = xfs_da_read_buf(state->args->trans,
+			error = xfs_da_node_read(state->args->trans,
 						state->args->dp,
 						blk->blkno, blk->disk_blkno,
 						&blk->bp, XFS_ATTR_FORK);
@@ -1726,7 +1715,7 @@ xfs_attr_refillstate(xfs_da_state_t *state)
 	ASSERT((path->active >= 0) && (path->active < XFS_DA_NODE_MAXDEPTH));
 	for (blk = path->blk, level = 0; level < path->active; blk++, level++) {
 		if (blk->disk_blkno) {
-			error = xfs_da_read_buf(state->args->trans,
+			error = xfs_da_node_read(state->args->trans,
 						state->args->dp,
 						blk->blkno, blk->disk_blkno,
 						&blk->bp, XFS_ATTR_FORK);
@@ -1755,6 +1744,8 @@ xfs_attr_node_get(xfs_da_args_t *args)
 	int error, retval;
 	int i;
 
+	trace_xfs_attr_node_get(args);
+
 	state = xfs_da_state_alloc();
 	state->args = args;
 	state->mp = args->dp->i_mount;
@@ -1804,6 +1795,8 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 	int error, i;
 	struct xfs_buf *bp;
 
+	trace_xfs_attr_node_list(context);
+
 	cursor = context->cursor;
 	cursor->initted = 1;
 
@@ -1814,7 +1807,7 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 	 */
 	bp = NULL;
 	if (cursor->blkno > 0) {
-		error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
+		error = xfs_da_node_read(NULL, context->dp, cursor->blkno, -1,
 					      &bp, XFS_ATTR_FORK);
 		if ((error != 0) && (error != EFSCORRUPTED))
 			return(error);
@@ -1856,17 +1849,11 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 	if (bp == NULL) {
 		cursor->blkno = 0;
 		for (;;) {
-			error = xfs_da_read_buf(NULL, context->dp,
+			error = xfs_da_node_read(NULL, context->dp,
 						      cursor->blkno, -1, &bp,
 						      XFS_ATTR_FORK);
 			if (error)
 				return(error);
-			if (unlikely(bp == NULL)) {
-				XFS_ERROR_REPORT("xfs_attr_node_list(2)",
-						 XFS_ERRLEVEL_LOW,
-						 context->dp->i_mount);
-				return(XFS_ERROR(EFSCORRUPTED));
-			}
 			node = bp->b_addr;
 			if (node->hdr.info.magic ==
 			    cpu_to_be16(XFS_ATTR_LEAF_MAGIC))
@@ -1907,14 +1894,6 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 	 */
 	for (;;) {
 		leaf = bp->b_addr;
-		if (unlikely(leaf->hdr.info.magic !=
-			     cpu_to_be16(XFS_ATTR_LEAF_MAGIC))) {
-			XFS_CORRUPTION_ERROR("xfs_attr_node_list(4)",
-					     XFS_ERRLEVEL_LOW,
-					     context->dp->i_mount, leaf);
-			xfs_trans_brelse(NULL, bp);
-			return(XFS_ERROR(EFSCORRUPTED));
-		}
 		error = xfs_attr_leaf_list_int(bp, context);
 		if (error) {
 			xfs_trans_brelse(NULL, bp);
@@ -1924,16 +1903,10 @@ xfs_attr_node_list(xfs_attr_list_context_t *context)
 			break;
 		cursor->blkno = be32_to_cpu(leaf->hdr.info.forw);
 		xfs_trans_brelse(NULL, bp);
-		error = xfs_da_read_buf(NULL, context->dp, cursor->blkno, -1,
-					      &bp, XFS_ATTR_FORK);
+		error = xfs_attr_leaf_read(NULL, context->dp, cursor->blkno, -1,
+					   &bp);
 		if (error)
-			return(error);
-		if (unlikely((bp == NULL))) {
-			XFS_ERROR_REPORT("xfs_attr_node_list(5)",
-					 XFS_ERRLEVEL_LOW,
-					 context->dp->i_mount);
-			return(XFS_ERROR(EFSCORRUPTED));
-		}
+			return error;
 	}
 	xfs_trans_brelse(NULL, bp);
 	return(0);
@@ -1959,6 +1932,8 @@ xfs_attr_rmtval_get(xfs_da_args_t *args)
 	int nmap, error, tmp, valuelen, blkcnt, i;
 	xfs_dablk_t lblkno;
 
+	trace_xfs_attr_rmtval_get(args);
+
 	ASSERT(!(args->flags & ATTR_KERNOVAL));
 
 	mp = args->dp->i_mount;
@@ -1980,7 +1955,7 @@ xfs_attr_rmtval_get(xfs_da_args_t *args)
 			dblkno = XFS_FSB_TO_DADDR(mp, map[i].br_startblock);
 			blkcnt = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
-						   dblkno, blkcnt, 0, &bp);
+						   dblkno, blkcnt, 0, &bp, NULL);
 			if (error)
 				return(error);
 
@@ -2014,6 +1989,8 @@ xfs_attr_rmtval_set(xfs_da_args_t *args)
 	xfs_dablk_t lblkno;
 	int blkcnt, valuelen, nmap, error, tmp, committed;
 
+	trace_xfs_attr_rmtval_set(args);
+
 	dp = args->dp;
 	mp = dp->i_mount;
 	src = args->value;
@@ -2143,6 +2120,8 @@ xfs_attr_rmtval_remove(xfs_da_args_t *args)
 	xfs_dablk_t lblkno;
 	int valuelen, blkcnt, nmap, error, done, committed;
 
+	trace_xfs_attr_rmtval_remove(args);
+
 	mp = args->dp->i_mount;
 
 	/*

fs/xfs/xfs_attr_leaf.c

@@ -57,7 +57,8 @@ STATIC int xfs_attr_leaf_create(xfs_da_args_t *args, xfs_dablk_t which_block,
 				struct xfs_buf **bpp);
 STATIC int xfs_attr_leaf_add_work(struct xfs_buf *leaf_buffer,
 				  xfs_da_args_t *args, int freemap_index);
-STATIC void xfs_attr_leaf_compact(xfs_trans_t *tp, struct xfs_buf *leaf_buffer);
+STATIC void xfs_attr_leaf_compact(struct xfs_da_args *args,
+				  struct xfs_buf *leaf_buffer);
 STATIC void xfs_attr_leaf_rebalance(xfs_da_state_t *state,
 						   xfs_da_state_blk_t *blk1,
 						   xfs_da_state_blk_t *blk2);
@@ -87,6 +88,52 @@ STATIC void xfs_attr_leaf_moveents(xfs_attr_leafblock_t *src_leaf,
 					 xfs_mount_t *mp);
 STATIC int xfs_attr_leaf_entsize(xfs_attr_leafblock_t *leaf, int index);
 
+static void
+xfs_attr_leaf_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_attr_leaf_hdr *hdr = bp->b_addr;
+	int			block_ok = 0;
+
+	block_ok = hdr->info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC);
+	if (!block_ok) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_attr_leaf_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_attr_leaf_verify(bp);
+}
+
+static void
+xfs_attr_leaf_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_attr_leaf_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_attr_leaf_buf_ops = {
+	.verify_read = xfs_attr_leaf_read_verify,
+	.verify_write = xfs_attr_leaf_write_verify,
+};
+
+int
+xfs_attr_leaf_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		bno,
+	xfs_daddr_t		mappedbno,
+	struct xfs_buf		**bpp)
+{
+	return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
+				XFS_ATTR_FORK, &xfs_attr_leaf_buf_ops);
+}
+
 /*========================================================================
  * Namespace helper routines
  *========================================================================*/
@@ -869,17 +916,16 @@ xfs_attr_leaf_to_node(xfs_da_args_t *args)
 	error = xfs_da_grow_inode(args, &blkno);
 	if (error)
 		goto out;
-	error = xfs_da_read_buf(args->trans, args->dp, 0, -1, &bp1,
-					     XFS_ATTR_FORK);
+	error = xfs_attr_leaf_read(args->trans, args->dp, 0, -1, &bp1);
 	if (error)
 		goto out;
-	ASSERT(bp1 != NULL);
+
 	bp2 = NULL;
 	error = xfs_da_get_buf(args->trans, args->dp, blkno, -1, &bp2,
 					    XFS_ATTR_FORK);
 	if (error)
 		goto out;
-	ASSERT(bp2 != NULL);
+	bp2->b_ops = bp1->b_ops;
 	memcpy(bp2->b_addr, bp1->b_addr, XFS_LBSIZE(dp->i_mount));
 	bp1 = NULL;
 	xfs_trans_log_buf(args->trans, bp2, 0, XFS_LBSIZE(dp->i_mount) - 1);
@@ -933,7 +979,7 @@ xfs_attr_leaf_create(
 					    XFS_ATTR_FORK);
 	if (error)
 		return(error);
-	ASSERT(bp != NULL);
+	bp->b_ops = &xfs_attr_leaf_buf_ops;
 	leaf = bp->b_addr;
 	memset((char *)leaf, 0, XFS_LBSIZE(dp->i_mount));
 	hdr = &leaf->hdr;
@@ -1071,7 +1117,7 @@ xfs_attr_leaf_add(
 	 * Compact the entries to coalesce free space.
 	 * This may change the hdr->count via dropping INCOMPLETE entries.
 	 */
-	xfs_attr_leaf_compact(args->trans, bp);
+	xfs_attr_leaf_compact(args, bp);
 
 	/*
 	 * After compaction, the block is guaranteed to have only one
@@ -1102,6 +1148,8 @@ xfs_attr_leaf_add_work(
 	xfs_mount_t *mp;
 	int tmp, i;
 
+	trace_xfs_attr_leaf_add_work(args);
+
 	leaf = bp->b_addr;
 	ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 	hdr = &leaf->hdr;
@@ -1214,15 +1262,17 @@ xfs_attr_leaf_add_work(
  */
 STATIC void
 xfs_attr_leaf_compact(
-	struct xfs_trans *trans,
-	struct xfs_buf	*bp)
+	struct xfs_da_args	*args,
+	struct xfs_buf		*bp)
 {
-	xfs_attr_leafblock_t *leaf_s, *leaf_d;
-	xfs_attr_leaf_hdr_t *hdr_s, *hdr_d;
-	xfs_mount_t *mp;
-	char *tmpbuffer;
+	xfs_attr_leafblock_t	*leaf_s, *leaf_d;
+	xfs_attr_leaf_hdr_t	*hdr_s, *hdr_d;
+	struct xfs_trans	*trans = args->trans;
+	struct xfs_mount	*mp = trans->t_mountp;
+	char			*tmpbuffer;
+
+	trace_xfs_attr_leaf_compact(args);
 
-	mp = trans->t_mountp;
 	tmpbuffer = kmem_alloc(XFS_LBSIZE(mp), KM_SLEEP);
 	ASSERT(tmpbuffer != NULL);
 	memcpy(tmpbuffer, bp->b_addr, XFS_LBSIZE(mp));
@@ -1345,9 +1395,8 @@ xfs_attr_leaf_rebalance(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
 		max  = be16_to_cpu(hdr2->firstused)
 						- sizeof(xfs_attr_leaf_hdr_t);
 		max -= be16_to_cpu(hdr2->count) * sizeof(xfs_attr_leaf_entry_t);
-		if (space > max) {
-			xfs_attr_leaf_compact(args->trans, blk2->bp);
-		}
+		if (space > max)
+			xfs_attr_leaf_compact(args, blk2->bp);
 
 		/*
 		 * Move high entries from leaf1 to low end of leaf2.
@@ -1378,9 +1427,8 @@ xfs_attr_leaf_rebalance(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
 		max  = be16_to_cpu(hdr1->firstused)
 						- sizeof(xfs_attr_leaf_hdr_t);
 		max -= be16_to_cpu(hdr1->count) * sizeof(xfs_attr_leaf_entry_t);
-		if (space > max) {
-			xfs_attr_leaf_compact(args->trans, blk1->bp);
-		}
+		if (space > max)
+			xfs_attr_leaf_compact(args, blk1->bp);
 
 		/*
 		 * Move low entries from leaf2 to high end of leaf1.
@@ -1577,6 +1625,8 @@ xfs_attr_leaf_toosmall(xfs_da_state_t *state, int *action)
 	xfs_dablk_t blkno;
 	struct xfs_buf *bp;
 
+	trace_xfs_attr_leaf_toosmall(state->args);
+
 	/*
 	 * Check for the degenerate case of the block being over 50% full.
 	 * If so, it's not worth even looking to see if we might be able
@@ -1636,18 +1686,16 @@ xfs_attr_leaf_toosmall(xfs_da_state_t *state, int *action)
 			blkno = be32_to_cpu(info->back);
 		if (blkno == 0)
 			continue;
-		error = xfs_da_read_buf(state->args->trans, state->args->dp,
-					blkno, -1, &bp, XFS_ATTR_FORK);
+		error = xfs_attr_leaf_read(state->args->trans, state->args->dp,
+					blkno, -1, &bp);
 		if (error)
 			return(error);
-		ASSERT(bp != NULL);
 
 		leaf = (xfs_attr_leafblock_t *)info;
 		count  = be16_to_cpu(leaf->hdr.count);
 		bytes  = state->blocksize - (state->blocksize>>2);
 		bytes -= be16_to_cpu(leaf->hdr.usedbytes);
 		leaf = bp->b_addr;
-		ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 		count += be16_to_cpu(leaf->hdr.count);
 		bytes -= be16_to_cpu(leaf->hdr.usedbytes);
 		bytes -= count * sizeof(xfs_attr_leaf_entry_t);
@@ -1702,6 +1750,8 @@ xfs_attr_leaf_remove(
 	int tablesize, tmp, i;
 	xfs_mount_t *mp;
 
+	trace_xfs_attr_leaf_remove(args);
+
 	leaf = bp->b_addr;
 	ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 	hdr = &leaf->hdr;
@@ -2511,15 +2561,11 @@ xfs_attr_leaf_clearflag(xfs_da_args_t *args)
 	/*
 	 * Set up the operation.
 	 */
-	error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
-					     XFS_ATTR_FORK);
-	if (error) {
+	error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
+	if (error)
 		return(error);
-	}
-	ASSERT(bp != NULL);
 
 	leaf = bp->b_addr;
-	ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 	ASSERT(args->index < be16_to_cpu(leaf->hdr.count));
 	ASSERT(args->index >= 0);
 	entry = &leaf->entries[ args->index ];
@@ -2576,15 +2622,11 @@ xfs_attr_leaf_setflag(xfs_da_args_t *args)
 	/*
 	 * Set up the operation.
 	 */
-	error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp,
-					     XFS_ATTR_FORK);
-	if (error) {
+	error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp);
+	if (error)
 		return(error);
-	}
-	ASSERT(bp != NULL);
 
 	leaf = bp->b_addr;
-	ASSERT(leaf->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 	ASSERT(args->index < be16_to_cpu(leaf->hdr.count));
 	ASSERT(args->index >= 0);
 	entry = &leaf->entries[ args->index ];
@@ -2633,35 +2675,28 @@ xfs_attr_leaf_flipflags(xfs_da_args_t *args)
 	/*
 	 * Read the block containing the "old" attr
 	 */
-	error = xfs_da_read_buf(args->trans, args->dp, args->blkno, -1, &bp1,
-					     XFS_ATTR_FORK);
-	if (error) {
-		return(error);
-	}
-	ASSERT(bp1 != NULL);
+	error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno, -1, &bp1);
+	if (error)
+		return error;
 
 	/*
 	 * Read the block containing the "new" attr, if it is different
 	 */
 	if (args->blkno2 != args->blkno) {
-		error = xfs_da_read_buf(args->trans, args->dp, args->blkno2,
-					-1, &bp2, XFS_ATTR_FORK);
-		if (error) {
-			return(error);
-		}
-		ASSERT(bp2 != NULL);
+		error = xfs_attr_leaf_read(args->trans, args->dp, args->blkno2,
+					   -1, &bp2);
+		if (error)
+			return error;
 	} else {
 		bp2 = bp1;
 	}
 
 	leaf1 = bp1->b_addr;
-	ASSERT(leaf1->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 	ASSERT(args->index < be16_to_cpu(leaf1->hdr.count));
 	ASSERT(args->index >= 0);
 	entry1 = &leaf1->entries[ args->index ];
 
 	leaf2 = bp2->b_addr;
-	ASSERT(leaf2->hdr.info.magic == cpu_to_be16(XFS_ATTR_LEAF_MAGIC));
 	ASSERT(args->index2 < be16_to_cpu(leaf2->hdr.count));
 	ASSERT(args->index2 >= 0);
 	entry2 = &leaf2->entries[ args->index2 ];
@@ -2746,7 +2781,7 @@ xfs_attr_root_inactive(xfs_trans_t **trans, xfs_inode_t *dp)
 	 * the extents in reverse order the extent containing
 	 * block 0 must still be there.
 	 */
-	error = xfs_da_read_buf(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK);
+	error = xfs_da_node_read(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK);
 	if (error)
 		return(error);
 	blkno = XFS_BUF_ADDR(bp);
@@ -2831,7 +2866,7 @@ xfs_attr_node_inactive(
 		 * traversal of the tree so we may deal with many blocks
 		 * before we come back to this one.
 		 */
-		error = xfs_da_read_buf(*trans, dp, child_fsb, -2, &child_bp,
+		error = xfs_da_node_read(*trans, dp, child_fsb, -2, &child_bp,
 						XFS_ATTR_FORK);
 		if (error)
 			return(error);
@@ -2872,8 +2907,8 @@ xfs_attr_node_inactive(
 		 * child block number.
 		 */
 		if ((i+1) < count) {
-			error = xfs_da_read_buf(*trans, dp, 0, parent_blkno,
-				&bp, XFS_ATTR_FORK);
+			error = xfs_da_node_read(*trans, dp, 0, parent_blkno,
+						 &bp, XFS_ATTR_FORK);
 			if (error)
 				return(error);
 			child_fsb = be32_to_cpu(node->btree[i+1].before);

fs/xfs/xfs_attr_leaf.h

@@ -261,4 +261,10 @@ int	xfs_attr_leaf_order(struct xfs_buf *leaf1_bp,
 				   struct xfs_buf *leaf2_bp);
 int	xfs_attr_leaf_newentsize(int namelen, int valuelen, int blocksize,
 					int *local);
+int	xfs_attr_leaf_read(struct xfs_trans *tp, struct xfs_inode *dp,
+			xfs_dablk_t bno, xfs_daddr_t mappedbno,
+			struct xfs_buf **bpp);
+
+extern const struct xfs_buf_ops xfs_attr_leaf_buf_ops;
+
 #endif	/* __XFS_ATTR_LEAF_H__ */

fs/xfs/xfs_bmap.c

@@ -2662,8 +2662,9 @@ xfs_bmap_btree_to_extents(
 	if ((error = xfs_btree_check_lptr(cur, cbno, 1)))
 		return error;
 #endif
-	if ((error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp,
-			XFS_BMAP_BTREE_REF)))
+	error = xfs_btree_read_bufl(mp, tp, cbno, 0, &cbp, XFS_BMAP_BTREE_REF,
+				&xfs_bmbt_buf_ops);
+	if (error)
 		return error;
 	cblock = XFS_BUF_TO_BLOCK(cbp);
 	if ((error = xfs_btree_check_block(cur, cblock, 0, cbp)))
@@ -3123,6 +3124,7 @@ xfs_bmap_extents_to_btree(
 	/*
 	 * Fill in the child block.
 	 */
+	abp->b_ops = &xfs_bmbt_buf_ops;
 	ablock = XFS_BUF_TO_BLOCK(abp);
 	ablock->bb_magic = cpu_to_be32(XFS_BMAP_MAGIC);
 	ablock->bb_level = 0;
@@ -3269,6 +3271,7 @@ xfs_bmap_local_to_extents(
 		ASSERT(args.len == 1);
 		*firstblock = args.fsbno;
 		bp = xfs_btree_get_bufl(args.mp, tp, args.fsbno, 0);
+		bp->b_ops = &xfs_bmbt_buf_ops;
 		memcpy(bp->b_addr, ifp->if_u1.if_data, ifp->if_bytes);
 		xfs_trans_log_buf(tp, bp, 0, ifp->if_bytes - 1);
 		xfs_bmap_forkoff_reset(args.mp, ip, whichfork);
@@ -4078,8 +4081,9 @@ xfs_bmap_read_extents(
 	 * pointer (leftmost) at each level.
 	 */
 	while (level-- > 0) {
-		if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
-				XFS_BMAP_BTREE_REF)))
+		error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
+				XFS_BMAP_BTREE_REF, &xfs_bmbt_buf_ops);
+		if (error)
 			return error;
 		block = XFS_BUF_TO_BLOCK(bp);
 		XFS_WANT_CORRUPTED_GOTO(
@@ -4124,7 +4128,8 @@ xfs_bmap_read_extents(
 		 */
 		nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 		if (nextbno != NULLFSBLOCK)
-			xfs_btree_reada_bufl(mp, nextbno, 1);
+			xfs_btree_reada_bufl(mp, nextbno, 1,
+					     &xfs_bmbt_buf_ops);
 		/*
 		 * Copy records into the extent records.
 		 */
@@ -4156,8 +4161,9 @@ xfs_bmap_read_extents(
 		 */
 		if (bno == NULLFSBLOCK)
 			break;
-		if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
-				XFS_BMAP_BTREE_REF)))
+		error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
+				XFS_BMAP_BTREE_REF, &xfs_bmbt_buf_ops);
+		if (error)
 			return error;
 		block = XFS_BUF_TO_BLOCK(bp);
 	}
@@ -5599,7 +5605,7 @@ xfs_getbmap(
 	xfs_ilock(ip, XFS_IOLOCK_SHARED);
 	if (whichfork == XFS_DATA_FORK && !(iflags & BMV_IF_DELALLOC)) {
 		if (ip->i_delayed_blks || XFS_ISIZE(ip) > ip->i_d.di_size) {
-			error = xfs_flush_pages(ip, 0, -1, 0, FI_REMAPF);
+			error = -filemap_write_and_wait(VFS_I(ip)->i_mapping);
 			if (error)
 				goto out_unlock_iolock;
 		}
@@ -5868,15 +5874,16 @@ xfs_bmap_check_leaf_extents(
 	 */
 	while (level-- > 0) {
 		/* See if buf is in cur first */
+		bp_release = 0;
 		bp = xfs_bmap_get_bp(cur, XFS_FSB_TO_DADDR(mp, bno));
-		if (bp) {
-			bp_release = 0;
-		} else {
+		if (!bp) {
 			bp_release = 1;
+			error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
+						XFS_BMAP_BTREE_REF,
+						&xfs_bmbt_buf_ops);
+			if (error)
+				goto error_norelse;
 		}
-		if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
-				XFS_BMAP_BTREE_REF)))
-			goto error_norelse;
 		block = XFS_BUF_TO_BLOCK(bp);
 		XFS_WANT_CORRUPTED_GOTO(
 			xfs_bmap_sanity_check(mp, bp, level),
@@ -5953,15 +5960,16 @@ xfs_bmap_check_leaf_extents(
 		if (bno == NULLFSBLOCK)
 			break;
 
+		bp_release = 0;
 		bp = xfs_bmap_get_bp(cur, XFS_FSB_TO_DADDR(mp, bno));
-		if (bp) {
-			bp_release = 0;
-		} else {
+		if (!bp) {
 			bp_release = 1;
+			error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
+						XFS_BMAP_BTREE_REF,
+						&xfs_bmbt_buf_ops);
+			if (error)
+				goto error_norelse;
 		}
-		if (!bp && (error = xfs_btree_read_bufl(mp, NULL, bno, 0, &bp,
-				XFS_BMAP_BTREE_REF)))
-			goto error_norelse;
 		block = XFS_BUF_TO_BLOCK(bp);
 	}
 	if (bp_release) {
@@ -6052,7 +6060,9 @@ xfs_bmap_count_tree(
 	struct xfs_btree_block	*block, *nextblock;
 	int			numrecs;
 
-	if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF)))
+	error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp, XFS_BMAP_BTREE_REF,
+						&xfs_bmbt_buf_ops);
+	if (error)
 		return error;
 	*count += 1;
 	block = XFS_BUF_TO_BLOCK(bp);
@@ -6061,8 +6071,10 @@ xfs_bmap_count_tree(
 		/* Not at node above leaves, count this level of nodes */
 		nextbno = be64_to_cpu(block->bb_u.l.bb_rightsib);
 		while (nextbno != NULLFSBLOCK) {
-			if ((error = xfs_btree_read_bufl(mp, tp, nextbno,
-				0, &nbp, XFS_BMAP_BTREE_REF)))
+			error = xfs_btree_read_bufl(mp, tp, nextbno, 0, &nbp,
+						XFS_BMAP_BTREE_REF,
+						&xfs_bmbt_buf_ops);
+			if (error)
 				return error;
 			*count += 1;
 			nextblock = XFS_BUF_TO_BLOCK(nbp);
@@ -6091,8 +6103,10 @@ xfs_bmap_count_tree(
 			if (nextbno == NULLFSBLOCK)
 				break;
 			bno = nextbno;
-			if ((error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
-				XFS_BMAP_BTREE_REF)))
+			error = xfs_btree_read_bufl(mp, tp, bno, 0, &bp,
+						XFS_BMAP_BTREE_REF,
+						&xfs_bmbt_buf_ops);
+			if (error)
 				return error;
 			*count += 1;
 			block = XFS_BUF_TO_BLOCK(bp);

fs/xfs/xfs_bmap_btree.c

@@ -36,6 +36,7 @@
 #include "xfs_bmap.h"
 #include "xfs_error.h"
 #include "xfs_quota.h"
+#include "xfs_trace.h"
 
 /*
  * Determine the extent state.
@@ -707,6 +708,67 @@ xfs_bmbt_key_diff(
 				      cur->bc_rec.b.br_startoff;
 }
 
+static void
+xfs_bmbt_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
+	unsigned int		level;
+	int			lblock_ok; /* block passes checks */
+
+	/* magic number and level verification.
+	 *
+	 * We don't know waht fork we belong to, so just verify that the level
+	 * is less than the maximum of the two. Later checks will be more
+	 * precise.
+	 */
+	level = be16_to_cpu(block->bb_level);
+	lblock_ok = block->bb_magic == cpu_to_be32(XFS_BMAP_MAGIC) &&
+		    level < max(mp->m_bm_maxlevels[0], mp->m_bm_maxlevels[1]);
+
+	/* numrecs verification */
+	lblock_ok = lblock_ok &&
+		be16_to_cpu(block->bb_numrecs) <= mp->m_bmap_dmxr[level != 0];
+
+	/* sibling pointer verification */
+	lblock_ok = lblock_ok &&
+		block->bb_u.l.bb_leftsib &&
+		(block->bb_u.l.bb_leftsib == cpu_to_be64(NULLDFSBNO) ||
+		 XFS_FSB_SANITY_CHECK(mp,
+			be64_to_cpu(block->bb_u.l.bb_leftsib))) &&
+		block->bb_u.l.bb_rightsib &&
+		(block->bb_u.l.bb_rightsib == cpu_to_be64(NULLDFSBNO) ||
+		 XFS_FSB_SANITY_CHECK(mp,
+			be64_to_cpu(block->bb_u.l.bb_rightsib)));
+
+	if (!lblock_ok) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_bmbt_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_bmbt_verify(bp);
+}
+
+static void
+xfs_bmbt_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_bmbt_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_bmbt_buf_ops = {
+	.verify_read = xfs_bmbt_read_verify,
+	.verify_write = xfs_bmbt_write_verify,
+};
+
+
 #ifdef DEBUG
 STATIC int
 xfs_bmbt_keys_inorder(
@@ -746,6 +808,7 @@ static const struct xfs_btree_ops xfs_bmbt_ops = {
 	.init_rec_from_cur	= xfs_bmbt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_bmbt_init_ptr_from_cur,
 	.key_diff		= xfs_bmbt_key_diff,
+	.buf_ops		= &xfs_bmbt_buf_ops,
 #ifdef DEBUG
 	.keys_inorder		= xfs_bmbt_keys_inorder,
 	.recs_inorder		= xfs_bmbt_recs_inorder,

fs/xfs/xfs_bmap_btree.h

@@ -236,5 +236,6 @@ extern int xfs_bmbt_maxrecs(struct xfs_mount *, int blocklen, int leaf);
 extern struct xfs_btree_cur *xfs_bmbt_init_cursor(struct xfs_mount *,
 		struct xfs_trans *, struct xfs_inode *, int);
 
+extern const struct xfs_buf_ops xfs_bmbt_buf_ops;
 
 #endif	/* __XFS_BMAP_BTREE_H__ */

fs/xfs/xfs_btree.c

@@ -266,9 +266,13 @@ xfs_btree_dup_cursor(
 	for (i = 0; i < new->bc_nlevels; i++) {
 		new->bc_ptrs[i] = cur->bc_ptrs[i];
 		new->bc_ra[i] = cur->bc_ra[i];
-		if ((bp = cur->bc_bufs[i])) {
-			if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
-				XFS_BUF_ADDR(bp), mp->m_bsize, 0, &bp))) {
+		bp = cur->bc_bufs[i];
+		if (bp) {
+			error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
+						   XFS_BUF_ADDR(bp), mp->m_bsize,
+						   0, &bp,
+						   cur->bc_ops->buf_ops);
+			if (error) {
 				xfs_btree_del_cursor(new, error);
 				*ncur = NULL;
 				return error;
@@ -609,25 +613,26 @@ xfs_btree_offsets(
  * Get a buffer for the block, return it read in.
  * Long-form addressing.
  */
-int					/* error */
+int
 xfs_btree_read_bufl(
-	xfs_mount_t	*mp,		/* file system mount point */
-	xfs_trans_t	*tp,		/* transaction pointer */
-	xfs_fsblock_t	fsbno,		/* file system block number */
-	uint		lock,		/* lock flags for read_buf */
-	xfs_buf_t	**bpp,		/* buffer for fsbno */
-	int		refval)		/* ref count value for buffer */
-{
-	xfs_buf_t	*bp;		/* return value */
+	struct xfs_mount	*mp,		/* file system mount point */
+	struct xfs_trans	*tp,		/* transaction pointer */
+	xfs_fsblock_t		fsbno,		/* file system block number */
+	uint			lock,		/* lock flags for read_buf */
+	struct xfs_buf		**bpp,		/* buffer for fsbno */
+	int			refval,		/* ref count value for buffer */
+	const struct xfs_buf_ops *ops)
+{
+	struct xfs_buf		*bp;		/* return value */
 	xfs_daddr_t		d;		/* real disk block address */
-	int		error;
+	int			error;
 
 	ASSERT(fsbno != NULLFSBLOCK);
 	d = XFS_FSB_TO_DADDR(mp, fsbno);
-	if ((error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
-			mp->m_bsize, lock, &bp))) {
+	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, d,
+				   mp->m_bsize, lock, &bp, ops);
+	if (error)
 		return error;
-	}
 	ASSERT(!xfs_buf_geterror(bp));
 	if (bp)
 		xfs_buf_set_ref(bp, refval);
@@ -642,15 +647,16 @@ xfs_btree_read_bufl(
 /* ARGSUSED */
 void
 xfs_btree_reada_bufl(
-	xfs_mount_t	*mp,		/* file system mount point */
-	xfs_fsblock_t	fsbno,		/* file system block number */
-	xfs_extlen_t	count)		/* count of filesystem blocks */
+	struct xfs_mount	*mp,		/* file system mount point */
+	xfs_fsblock_t		fsbno,		/* file system block number */
+	xfs_extlen_t		count,		/* count of filesystem blocks */
+	const struct xfs_buf_ops *ops)
 {
 	xfs_daddr_t		d;
 
 	ASSERT(fsbno != NULLFSBLOCK);
 	d = XFS_FSB_TO_DADDR(mp, fsbno);
-	xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
+	xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, ops);
 }
 
 /*
@@ -660,17 +666,18 @@ xfs_btree_reada_bufl(
 /* ARGSUSED */
 void
 xfs_btree_reada_bufs(
-	xfs_mount_t	*mp,		/* file system mount point */
-	xfs_agnumber_t	agno,		/* allocation group number */
-	xfs_agblock_t	agbno,		/* allocation group block number */
-	xfs_extlen_t	count)		/* count of filesystem blocks */
+	struct xfs_mount	*mp,		/* file system mount point */
+	xfs_agnumber_t		agno,		/* allocation group number */
+	xfs_agblock_t		agbno,		/* allocation group block number */
+	xfs_extlen_t		count,		/* count of filesystem blocks */
+	const struct xfs_buf_ops *ops)
 {
 	xfs_daddr_t		d;
 
 	ASSERT(agno != NULLAGNUMBER);
 	ASSERT(agbno != NULLAGBLOCK);
 	d = XFS_AGB_TO_DADDR(mp, agno, agbno);
-	xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count);
+	xfs_buf_readahead(mp->m_ddev_targp, d, mp->m_bsize * count, ops);
 }
 
 STATIC int
@@ -684,12 +691,14 @@ xfs_btree_readahead_lblock(
 	xfs_dfsbno_t		right = be64_to_cpu(block->bb_u.l.bb_rightsib);
 
 	if ((lr & XFS_BTCUR_LEFTRA) && left != NULLDFSBNO) {
-		xfs_btree_reada_bufl(cur->bc_mp, left, 1);
+		xfs_btree_reada_bufl(cur->bc_mp, left, 1,
+				     cur->bc_ops->buf_ops);
 		rval++;
 	}
 
 	if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLDFSBNO) {
-		xfs_btree_reada_bufl(cur->bc_mp, right, 1);
+		xfs_btree_reada_bufl(cur->bc_mp, right, 1,
+				     cur->bc_ops->buf_ops);
 		rval++;
 	}
 
@@ -709,13 +718,13 @@ xfs_btree_readahead_sblock(
 
 	if ((lr & XFS_BTCUR_LEFTRA) && left != NULLAGBLOCK) {
 		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
-				     left, 1);
+				     left, 1, cur->bc_ops->buf_ops);
 		rval++;
 	}
 
 	if ((lr & XFS_BTCUR_RIGHTRA) && right != NULLAGBLOCK) {
 		xfs_btree_reada_bufs(cur->bc_mp, cur->bc_private.a.agno,
-				     right, 1);
+				     right, 1, cur->bc_ops->buf_ops);
 		rval++;
 	}
 
@@ -853,18 +862,22 @@ xfs_btree_set_sibling(
 	}
 }
 
-STATIC void
+void
 xfs_btree_init_block(
-	struct xfs_btree_cur	*cur,
-	int			level,
-	int			numrecs,
-	struct xfs_btree_block	*new)	/* new block */
+	struct xfs_mount *mp,
+	struct xfs_buf	*bp,
+	__u32		magic,
+	__u16		level,
+	__u16		numrecs,
+	unsigned int	flags)
 {
-	new->bb_magic = cpu_to_be32(xfs_magics[cur->bc_btnum]);
+	struct xfs_btree_block	*new = XFS_BUF_TO_BLOCK(bp);
+
+	new->bb_magic = cpu_to_be32(magic);
 	new->bb_level = cpu_to_be16(level);
 	new->bb_numrecs = cpu_to_be16(numrecs);
 
-	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
+	if (flags & XFS_BTREE_LONG_PTRS) {
 		new->bb_u.l.bb_leftsib = cpu_to_be64(NULLDFSBNO);
 		new->bb_u.l.bb_rightsib = cpu_to_be64(NULLDFSBNO);
 	} else {
@@ -873,6 +886,17 @@ xfs_btree_init_block(
 	}
 }
 
+STATIC void
+xfs_btree_init_block_cur(
+	struct xfs_btree_cur	*cur,
+	int			level,
+	int			numrecs,
+	struct xfs_buf		*bp)
+{
+	xfs_btree_init_block(cur->bc_mp, bp, xfs_magics[cur->bc_btnum],
+			       level, numrecs, cur->bc_flags);
+}
+
 /*
  * Return true if ptr is the last record in the btree and
  * we need to track updateѕ to this record.  The decision
@@ -972,6 +996,7 @@ xfs_btree_get_buf_block(
 	if (!*bpp)
 		return ENOMEM;
 
+	(*bpp)->b_ops = cur->bc_ops->buf_ops;
 	*block = XFS_BUF_TO_BLOCK(*bpp);
 	return 0;
 }
@@ -998,19 +1023,15 @@ xfs_btree_read_buf_block(
 
 	d = xfs_btree_ptr_to_daddr(cur, ptr);
 	error = xfs_trans_read_buf(mp, cur->bc_tp, mp->m_ddev_targp, d,
-				   mp->m_bsize, flags, bpp);
+				   mp->m_bsize, flags, bpp,
+				   cur->bc_ops->buf_ops);
 	if (error)
 		return error;
 
 	ASSERT(!xfs_buf_geterror(*bpp));
-
 	xfs_btree_set_refs(cur, *bpp);
 	*block = XFS_BUF_TO_BLOCK(*bpp);
-
-	error = xfs_btree_check_block(cur, *block, level, *bpp);
-	if (error)
-		xfs_trans_brelse(cur->bc_tp, *bpp);
-	return error;
+	return 0;
 }
 
 /*
@@ -2183,7 +2204,7 @@ xfs_btree_split(
 		goto error0;
 
 	/* Fill in the btree header for the new right block. */
-	xfs_btree_init_block(cur, xfs_btree_get_level(left), 0, right);
+	xfs_btree_init_block_cur(cur, xfs_btree_get_level(left), 0, rbp);
 
 	/*
 	 * Split the entries between the old and the new block evenly.
@@ -2492,7 +2513,7 @@ xfs_btree_new_root(
 		nptr = 2;
 	}
 	/* Fill in the new block's btree header and log it. */
-	xfs_btree_init_block(cur, cur->bc_nlevels, 2, new);
+	xfs_btree_init_block_cur(cur, cur->bc_nlevels, 2, nbp);
 	xfs_btree_log_block(cur, nbp, XFS_BB_ALL_BITS);
 	ASSERT(!xfs_btree_ptr_is_null(cur, &lptr) &&
 			!xfs_btree_ptr_is_null(cur, &rptr));

fs/xfs/xfs_btree.h

@@ -188,6 +188,8 @@ struct xfs_btree_ops {
 	__int64_t (*key_diff)(struct xfs_btree_cur *cur,
 			      union xfs_btree_key *key);
 
+	const struct xfs_buf_ops	*buf_ops;
+
 #ifdef DEBUG
 	/* check that k1 is lower than k2 */
 	int	(*keys_inorder)(struct xfs_btree_cur *cur,
@@ -355,7 +357,8 @@ xfs_btree_read_bufl(
 	xfs_fsblock_t		fsbno,	/* file system block number */
 	uint			lock,	/* lock flags for read_buf */
 	struct xfs_buf		**bpp,	/* buffer for fsbno */
-	int			refval);/* ref count value for buffer */
+	int			refval,	/* ref count value for buffer */
+	const struct xfs_buf_ops *ops);
 
 /*
  * Read-ahead the block, don't wait for it, don't return a buffer.
@@ -365,7 +368,8 @@ void					/* error */
 xfs_btree_reada_bufl(
 	struct xfs_mount	*mp,	/* file system mount point */
 	xfs_fsblock_t		fsbno,	/* file system block number */
-	xfs_extlen_t		count);	/* count of filesystem blocks */
+	xfs_extlen_t		count,	/* count of filesystem blocks */
+	const struct xfs_buf_ops *ops);
 
 /*
  * Read-ahead the block, don't wait for it, don't return a buffer.
@@ -376,8 +380,20 @@ xfs_btree_reada_bufs(
 	struct xfs_mount	*mp,	/* file system mount point */
 	xfs_agnumber_t		agno,	/* allocation group number */
 	xfs_agblock_t		agbno,	/* allocation group block number */
-	xfs_extlen_t		count);	/* count of filesystem blocks */
+	xfs_extlen_t		count,	/* count of filesystem blocks */
+	const struct xfs_buf_ops *ops);
 
+/*
+ * Initialise a new btree block header
+ */
+void
+xfs_btree_init_block(
+	struct xfs_mount *mp,
+	struct xfs_buf	*bp,
+	__u32		magic,
+	__u16		level,
+	__u16		numrecs,
+	unsigned int	flags);
 
 /*
  * Common btree core entry points.

fs/xfs/xfs_buf.c

@@ -569,7 +569,9 @@ found:
 	 */
 	if (bp->b_flags & XBF_STALE) {
 		ASSERT((bp->b_flags & _XBF_DELWRI_Q) == 0);
+		ASSERT(bp->b_iodone == NULL);
 		bp->b_flags &= _XBF_KMEM | _XBF_PAGES;
+		bp->b_ops = NULL;
 	}
 
 	trace_xfs_buf_find(bp, flags, _RET_IP_);
@@ -654,7 +656,8 @@ xfs_buf_read_map(
 	struct xfs_buftarg	*target,
 	struct xfs_buf_map	*map,
 	int			nmaps,
-	xfs_buf_flags_t		flags)
+	xfs_buf_flags_t		flags,
+	const struct xfs_buf_ops *ops)
 {
 	struct xfs_buf		*bp;
 
@@ -666,6 +669,7 @@ xfs_buf_read_map(
 
 		if (!XFS_BUF_ISDONE(bp)) {
 			XFS_STATS_INC(xb_get_read);
+			bp->b_ops = ops;
 			_xfs_buf_read(bp, flags);
 		} else if (flags & XBF_ASYNC) {
 			/*
@@ -691,13 +695,14 @@ void
 xfs_buf_readahead_map(
 	struct xfs_buftarg	*target,
 	struct xfs_buf_map	*map,
-	int			nmaps)
+	int			nmaps,
+	const struct xfs_buf_ops *ops)
 {
 	if (bdi_read_congested(target->bt_bdi))
 		return;
 
 	xfs_buf_read_map(target, map, nmaps,
-		     XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD);
+		     XBF_TRYLOCK|XBF_ASYNC|XBF_READ_AHEAD, ops);
 }
 
 /*
@@ -709,10 +714,10 @@ xfs_buf_read_uncached(
 	struct xfs_buftarg	*target,
 	xfs_daddr_t		daddr,
 	size_t			numblks,
-	int			flags)
+	int			flags,
+	const struct xfs_buf_ops *ops)
 {
-	xfs_buf_t		*bp;
-	int			error;
+	struct xfs_buf		*bp;
 
 	bp = xfs_buf_get_uncached(target, numblks, flags);
 	if (!bp)
@@ -723,13 +728,10 @@ xfs_buf_read_uncached(
 	bp->b_bn = daddr;
 	bp->b_maps[0].bm_bn = daddr;
 	bp->b_flags |= XBF_READ;
+	bp->b_ops = ops;
 
 	xfsbdstrat(target->bt_mount, bp);
-	error = xfs_buf_iowait(bp);
-	if (error) {
-		xfs_buf_relse(bp);
-		return NULL;
-	}
+	xfs_buf_iowait(bp);
 	return bp;
 }
 
@@ -999,27 +1001,37 @@ STATIC void
 xfs_buf_iodone_work(
 	struct work_struct	*work)
 {
-	xfs_buf_t		*bp =
+	struct xfs_buf		*bp =
 		container_of(work, xfs_buf_t, b_iodone_work);
+	bool			read = !!(bp->b_flags & XBF_READ);
+
+	bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_READ_AHEAD);
+	if (read && bp->b_ops)
+		bp->b_ops->verify_read(bp);
 
 	if (bp->b_iodone)
 		(*(bp->b_iodone))(bp);
 	else if (bp->b_flags & XBF_ASYNC)
 		xfs_buf_relse(bp);
+	else {
+		ASSERT(read && bp->b_ops);
+		complete(&bp->b_iowait);
+	}
 }
 
 void
 xfs_buf_ioend(
-	xfs_buf_t		*bp,
-	int			schedule)
+	struct xfs_buf	*bp,
+	int		schedule)
 {
+	bool		read = !!(bp->b_flags & XBF_READ);
+
 	trace_xfs_buf_iodone(bp, _RET_IP_);
 
-	bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_READ_AHEAD);
 	if (bp->b_error == 0)
 		bp->b_flags |= XBF_DONE;
 
-	if ((bp->b_iodone) || (bp->b_flags & XBF_ASYNC)) {
+	if (bp->b_iodone || (read && bp->b_ops) || (bp->b_flags & XBF_ASYNC)) {
 		if (schedule) {
 			INIT_WORK(&bp->b_iodone_work, xfs_buf_iodone_work);
 			queue_work(xfslogd_workqueue, &bp->b_iodone_work);
@@ -1027,6 +1039,7 @@ xfs_buf_ioend(
 			xfs_buf_iodone_work(&bp->b_iodone_work);
 		}
 	} else {
+		bp->b_flags &= ~(XBF_READ | XBF_WRITE | XBF_READ_AHEAD);
 		complete(&bp->b_iowait);
 	}
 }
@@ -1314,6 +1327,20 @@ _xfs_buf_ioapply(
 			rw |= REQ_FUA;
 		if (bp->b_flags & XBF_FLUSH)
 			rw |= REQ_FLUSH;
+
+		/*
+		 * Run the write verifier callback function if it exists. If
+		 * this function fails it will mark the buffer with an error and
+		 * the IO should not be dispatched.
+		 */
+		if (bp->b_ops) {
+			bp->b_ops->verify_write(bp);
+			if (bp->b_error) {
+				xfs_force_shutdown(bp->b_target->bt_mount,
+						   SHUTDOWN_CORRUPT_INCORE);
+				return;
+			}
+		}
 	} else if (bp->b_flags & XBF_READ_AHEAD) {
 		rw = READA;
 	} else {

fs/xfs/xfs_buf.h

@@ -100,6 +100,7 @@ typedef struct xfs_buftarg {
 struct xfs_buf;
 typedef void (*xfs_buf_iodone_t)(struct xfs_buf *);
 
+
 #define XB_PAGES	2
 
 struct xfs_buf_map {
@@ -110,6 +111,11 @@ struct xfs_buf_map {
 #define DEFINE_SINGLE_BUF_MAP(map, blkno, numblk) \
 	struct xfs_buf_map (map) = { .bm_bn = (blkno), .bm_len = (numblk) };
 
+struct xfs_buf_ops {
+	void (*verify_read)(struct xfs_buf *);
+	void (*verify_write)(struct xfs_buf *);
+};
+
 typedef struct xfs_buf {
 	/*
 	 * first cacheline holds all the fields needed for an uncontended cache
@@ -153,13 +159,13 @@ typedef struct xfs_buf {
 	unsigned int		b_page_count;	/* size of page array */
 	unsigned int		b_offset;	/* page offset in first page */
 	unsigned short		b_error;	/* error code on I/O */
+	const struct xfs_buf_ops	*b_ops;
 
 #ifdef XFS_BUF_LOCK_TRACKING
 	int			b_last_holder;
 #endif
 } xfs_buf_t;
 
-
 /* Finding and Reading Buffers */
 struct xfs_buf *_xfs_buf_find(struct xfs_buftarg *target,
 			      struct xfs_buf_map *map, int nmaps,
@@ -196,9 +202,11 @@ struct xfs_buf *xfs_buf_get_map(struct xfs_buftarg *target,
 			       xfs_buf_flags_t flags);
 struct xfs_buf *xfs_buf_read_map(struct xfs_buftarg *target,
 			       struct xfs_buf_map *map, int nmaps,
-			       xfs_buf_flags_t flags);
+			       xfs_buf_flags_t flags,
+			       const struct xfs_buf_ops *ops);
 void xfs_buf_readahead_map(struct xfs_buftarg *target,
-			       struct xfs_buf_map *map, int nmaps);
+			       struct xfs_buf_map *map, int nmaps,
+			       const struct xfs_buf_ops *ops);
 
 static inline struct xfs_buf *
 xfs_buf_get(
@@ -216,20 +224,22 @@ xfs_buf_read(
 	struct xfs_buftarg	*target,
 	xfs_daddr_t		blkno,
 	size_t			numblks,
-	xfs_buf_flags_t		flags)
+	xfs_buf_flags_t		flags,
+	const struct xfs_buf_ops *ops)
 {
 	DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
-	return xfs_buf_read_map(target, &map, 1, flags);
+	return xfs_buf_read_map(target, &map, 1, flags, ops);
 }
 
 static inline void
 xfs_buf_readahead(
 	struct xfs_buftarg	*target,
 	xfs_daddr_t		blkno,
-	size_t			numblks)
+	size_t			numblks,
+	const struct xfs_buf_ops *ops)
 {
 	DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
-	return xfs_buf_readahead_map(target, &map, 1);
+	return xfs_buf_readahead_map(target, &map, 1, ops);
 }
 
 struct xfs_buf *xfs_buf_get_empty(struct xfs_buftarg *target, size_t numblks);
@@ -239,7 +249,8 @@ int xfs_buf_associate_memory(struct xfs_buf *bp, void *mem, size_t length);
 struct xfs_buf *xfs_buf_get_uncached(struct xfs_buftarg *target, size_t numblks,
 				int flags);
 struct xfs_buf *xfs_buf_read_uncached(struct xfs_buftarg *target,
-				xfs_daddr_t daddr, size_t numblks, int flags);
+				xfs_daddr_t daddr, size_t numblks, int flags,
+				const struct xfs_buf_ops *ops);
 void xfs_buf_hold(struct xfs_buf *bp);
 
 /* Releasing Buffers */

fs/xfs/xfs_cksum.h

@@ -0,0 +1,63 @@
+#ifndef _XFS_CKSUM_H
+#define _XFS_CKSUM_H 1
+
+#define XFS_CRC_SEED	(~(__uint32_t)0)
+
+/*
+ * Calculate the intermediate checksum for a buffer that has the CRC field
+ * inside it.  The offset of the 32bit crc fields is passed as the
+ * cksum_offset parameter.
+ */
+static inline __uint32_t
+xfs_start_cksum(char *buffer, size_t length, unsigned long cksum_offset)
+{
+	__uint32_t zero = 0;
+	__uint32_t crc;
+
+	/* Calculate CRC up to the checksum. */
+	crc = crc32c(XFS_CRC_SEED, buffer, cksum_offset);
+
+	/* Skip checksum field */
+	crc = crc32c(crc, &zero, sizeof(__u32));
+
+	/* Calculate the rest of the CRC. */
+	return crc32c(crc, &buffer[cksum_offset + sizeof(__be32)],
+		      length - (cksum_offset + sizeof(__be32)));
+}
+
+/*
+ * Convert the intermediate checksum to the final ondisk format.
+ *
+ * The CRC32c calculation uses LE format even on BE machines, but returns the
+ * result in host endian format. Hence we need to byte swap it back to LE format
+ * so that it is consistent on disk.
+ */
+static inline __le32
+xfs_end_cksum(__uint32_t crc)
+{
+	return ~cpu_to_le32(crc);
+}
+
+/*
+ * Helper to generate the checksum for a buffer.
+ */
+static inline void
+xfs_update_cksum(char *buffer, size_t length, unsigned long cksum_offset)
+{
+	__uint32_t crc = xfs_start_cksum(buffer, length, cksum_offset);
+
+	*(__le32 *)(buffer + cksum_offset) = xfs_end_cksum(crc);
+}
+
+/*
+ * Helper to verify the checksum for a buffer.
+ */
+static inline int
+xfs_verify_cksum(char *buffer, size_t length, unsigned long cksum_offset)
+{
+	__uint32_t crc = xfs_start_cksum(buffer, length, cksum_offset);
+
+	return *(__le32 *)(buffer + cksum_offset) == xfs_end_cksum(crc);
+}
+
+#endif /* _XFS_CKSUM_H */

fs/xfs/xfs_da_btree.c

@@ -91,6 +91,84 @@ STATIC int	xfs_da_blk_unlink(xfs_da_state_t *state,
 				  xfs_da_state_blk_t *save_blk);
 STATIC void	xfs_da_state_kill_altpath(xfs_da_state_t *state);
 
+static void
+xfs_da_node_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_da_node_hdr *hdr = bp->b_addr;
+	int			block_ok = 0;
+
+	block_ok = hdr->info.magic == cpu_to_be16(XFS_DA_NODE_MAGIC);
+	block_ok = block_ok &&
+			be16_to_cpu(hdr->level) > 0 &&
+			be16_to_cpu(hdr->count) > 0 ;
+	if (!block_ok) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+
+}
+
+static void
+xfs_da_node_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_da_node_verify(bp);
+}
+
+/*
+ * leaf/node format detection on trees is sketchy, so a node read can be done on
+ * leaf level blocks when detection identifies the tree as a node format tree
+ * incorrectly. In this case, we need to swap the verifier to match the correct
+ * format of the block being read.
+ */
+static void
+xfs_da_node_read_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_da_blkinfo	*info = bp->b_addr;
+
+	switch (be16_to_cpu(info->magic)) {
+		case XFS_DA_NODE_MAGIC:
+			xfs_da_node_verify(bp);
+			break;
+		case XFS_ATTR_LEAF_MAGIC:
+			bp->b_ops = &xfs_attr_leaf_buf_ops;
+			bp->b_ops->verify_read(bp);
+			return;
+		case XFS_DIR2_LEAFN_MAGIC:
+			bp->b_ops = &xfs_dir2_leafn_buf_ops;
+			bp->b_ops->verify_read(bp);
+			return;
+		default:
+			XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW,
+					     mp, info);
+			xfs_buf_ioerror(bp, EFSCORRUPTED);
+			break;
+	}
+}
+
+const struct xfs_buf_ops xfs_da_node_buf_ops = {
+	.verify_read = xfs_da_node_read_verify,
+	.verify_write = xfs_da_node_write_verify,
+};
+
+
+int
+xfs_da_node_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		bno,
+	xfs_daddr_t		mappedbno,
+	struct xfs_buf		**bpp,
+	int			which_fork)
+{
+	return xfs_da_read_buf(tp, dp, bno, mappedbno, bpp,
+					which_fork, &xfs_da_node_buf_ops);
+}
+
 /*========================================================================
  * Routines used for growing the Btree.
  *========================================================================*/
@@ -125,6 +203,7 @@ xfs_da_node_create(xfs_da_args_t *args, xfs_dablk_t blkno, int level,
 	xfs_trans_log_buf(tp, bp,
 		XFS_DA_LOGRANGE(node, &node->hdr, sizeof(node->hdr)));
 
+	bp->b_ops = &xfs_da_node_buf_ops;
 	*bpp = bp;
 	return(0);
 }
@@ -324,6 +403,8 @@ xfs_da_root_split(xfs_da_state_t *state, xfs_da_state_blk_t *blk1,
 	}
 	memcpy(node, oldroot, size);
 	xfs_trans_log_buf(tp, bp, 0, size - 1);
+
+	bp->b_ops = blk1->bp->b_ops;
 	blk1->bp = bp;
 	blk1->blkno = blkno;
 
@@ -746,7 +827,7 @@ xfs_da_root_join(xfs_da_state_t *state, xfs_da_state_blk_t *root_blk)
 	 */
 	child = be32_to_cpu(oldroot->btree[0].before);
 	ASSERT(child != 0);
-	error = xfs_da_read_buf(args->trans, args->dp, child, -1, &bp,
+	error = xfs_da_node_read(args->trans, args->dp, child, -1, &bp,
 					     args->whichfork);
 	if (error)
 		return(error);
@@ -754,7 +835,14 @@ xfs_da_root_join(xfs_da_state_t *state, xfs_da_state_blk_t *root_blk)
 	xfs_da_blkinfo_onlychild_validate(bp->b_addr,
 					be16_to_cpu(oldroot->hdr.level));
 
+	/*
+	 * This could be copying a leaf back into the root block in the case of
+	 * there only being a single leaf block left in the tree. Hence we have
+	 * to update the b_ops pointer as well to match the buffer type change
+	 * that could occur.
+	 */
 	memcpy(root_blk->bp->b_addr, bp->b_addr, state->blocksize);
+	root_blk->bp->b_ops = bp->b_ops;
 	xfs_trans_log_buf(args->trans, root_blk->bp, 0, state->blocksize - 1);
 	error = xfs_da_shrink_inode(args, child, bp);
 	return(error);
@@ -779,6 +867,8 @@ xfs_da_node_toosmall(xfs_da_state_t *state, int *action)
 	xfs_dablk_t blkno;
 	struct xfs_buf *bp;
 
+	trace_xfs_da_node_toosmall(state->args);
+
 	/*
 	 * Check for the degenerate case of the block being over 50% full.
 	 * If so, it's not worth even looking to see if we might be able
@@ -835,7 +925,7 @@ xfs_da_node_toosmall(xfs_da_state_t *state, int *action)
 			blkno = be32_to_cpu(info->back);
 		if (blkno == 0)
 			continue;
-		error = xfs_da_read_buf(state->args->trans, state->args->dp,
+		error = xfs_da_node_read(state->args->trans, state->args->dp,
 					blkno, -1, &bp, state->args->whichfork);
 		if (error)
 			return(error);
@@ -900,6 +990,8 @@ xfs_da_fixhashpath(xfs_da_state_t *state, xfs_da_state_path_t *path)
 	xfs_dahash_t lasthash=0;
 	int level, count;
 
+	trace_xfs_da_fixhashpath(state->args);
+
 	level = path->active-1;
 	blk = &path->blk[ level ];
 	switch (blk->magic) {
@@ -1079,7 +1171,7 @@ xfs_da_node_lookup_int(xfs_da_state_t *state, int *result)
 		 * Read the next node down in the tree.
 		 */
 		blk->blkno = blkno;
-		error = xfs_da_read_buf(args->trans, args->dp, blkno,
+		error = xfs_da_node_read(args->trans, args->dp, blkno,
 					-1, &blk->bp, args->whichfork);
 		if (error) {
 			blk->blkno = 0;
@@ -1241,7 +1333,7 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
 		new_info->forw = cpu_to_be32(old_blk->blkno);
 		new_info->back = old_info->back;
 		if (old_info->back) {
-			error = xfs_da_read_buf(args->trans, args->dp,
+			error = xfs_da_node_read(args->trans, args->dp,
 						be32_to_cpu(old_info->back),
 						-1, &bp, args->whichfork);
 			if (error)
@@ -1262,7 +1354,7 @@ xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
 		new_info->forw = old_info->forw;
 		new_info->back = cpu_to_be32(old_blk->blkno);
 		if (old_info->forw) {
-			error = xfs_da_read_buf(args->trans, args->dp,
+			error = xfs_da_node_read(args->trans, args->dp,
 						be32_to_cpu(old_info->forw),
 						-1, &bp, args->whichfork);
 			if (error)
@@ -1362,7 +1454,7 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
 		trace_xfs_da_unlink_back(args);
 		save_info->back = drop_info->back;
 		if (drop_info->back) {
-			error = xfs_da_read_buf(args->trans, args->dp,
+			error = xfs_da_node_read(args->trans, args->dp,
 						be32_to_cpu(drop_info->back),
 						-1, &bp, args->whichfork);
 			if (error)
@@ -1379,7 +1471,7 @@ xfs_da_blk_unlink(xfs_da_state_t *state, xfs_da_state_blk_t *drop_blk,
 		trace_xfs_da_unlink_forward(args);
 		save_info->forw = drop_info->forw;
 		if (drop_info->forw) {
-			error = xfs_da_read_buf(args->trans, args->dp,
+			error = xfs_da_node_read(args->trans, args->dp,
 						be32_to_cpu(drop_info->forw),
 						-1, &bp, args->whichfork);
 			if (error)
@@ -1417,6 +1509,8 @@ xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
 	xfs_dablk_t blkno=0;
 	int level, error;
 
+	trace_xfs_da_path_shift(state->args);
+
 	/*
 	 * Roll up the Btree looking for the first block where our
 	 * current index is not at the edge of the block.  Note that
@@ -1463,8 +1557,8 @@ xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
 		 * Read the next child block.
 		 */
 		blk->blkno = blkno;
-		error = xfs_da_read_buf(args->trans, args->dp, blkno, -1,
-						     &blk->bp, args->whichfork);
+		error = xfs_da_node_read(args->trans, args->dp, blkno, -1,
+					&blk->bp, args->whichfork);
 		if (error)
 			return(error);
 		ASSERT(blk->bp != NULL);
@@ -1727,7 +1821,8 @@ xfs_da_swap_lastblock(
 	 * Read the last block in the btree space.
 	 */
 	last_blkno = (xfs_dablk_t)lastoff - mp->m_dirblkfsbs;
-	if ((error = xfs_da_read_buf(tp, ip, last_blkno, -1, &last_buf, w)))
+	error = xfs_da_node_read(tp, ip, last_blkno, -1, &last_buf, w);
+	if (error)
 		return error;
 	/*
 	 * Copy the last block into the dead buffer and log it.
@@ -1753,7 +1848,8 @@ xfs_da_swap_lastblock(
 	 * If the moved block has a left sibling, fix up the pointers.
 	 */
 	if ((sib_blkno = be32_to_cpu(dead_info->back))) {
-		if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
+		error = xfs_da_node_read(tp, ip, sib_blkno, -1, &sib_buf, w);
+		if (error)
 			goto done;
 		sib_info = sib_buf->b_addr;
 		if (unlikely(
@@ -1774,7 +1870,8 @@ xfs_da_swap_lastblock(
 	 * If the moved block has a right sibling, fix up the pointers.
 	 */
 	if ((sib_blkno = be32_to_cpu(dead_info->forw))) {
-		if ((error = xfs_da_read_buf(tp, ip, sib_blkno, -1, &sib_buf, w)))
+		error = xfs_da_node_read(tp, ip, sib_blkno, -1, &sib_buf, w);
+		if (error)
 			goto done;
 		sib_info = sib_buf->b_addr;
 		if (unlikely(
@@ -1797,7 +1894,8 @@ xfs_da_swap_lastblock(
 	 * Walk down the tree looking for the parent of the moved block.
 	 */
 	for (;;) {
-		if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
+		error = xfs_da_node_read(tp, ip, par_blkno, -1, &par_buf, w);
+		if (error)
 			goto done;
 		par_node = par_buf->b_addr;
 		if (unlikely(par_node->hdr.info.magic !=
@@ -1847,7 +1945,8 @@ xfs_da_swap_lastblock(
 			error = XFS_ERROR(EFSCORRUPTED);
 			goto done;
 		}
-		if ((error = xfs_da_read_buf(tp, ip, par_blkno, -1, &par_buf, w)))
+		error = xfs_da_node_read(tp, ip, par_blkno, -1, &par_buf, w);
+		if (error)
 			goto done;
 		par_node = par_buf->b_addr;
 		if (unlikely(
@@ -2133,7 +2232,8 @@ xfs_da_read_buf(
 	xfs_dablk_t		bno,
 	xfs_daddr_t		mappedbno,
 	struct xfs_buf		**bpp,
-	int			whichfork)
+	int			whichfork,
+	const struct xfs_buf_ops *ops)
 {
 	struct xfs_buf		*bp;
 	struct xfs_buf_map	map;
@@ -2155,7 +2255,7 @@ xfs_da_read_buf(
 
 	error = xfs_trans_read_buf_map(dp->i_mount, trans,
 					dp->i_mount->m_ddev_targp,
-					mapp, nmap, 0, &bp);
+					mapp, nmap, 0, &bp, ops);
 	if (error)
 		goto out_free;
 
@@ -2211,9 +2311,10 @@ xfs_da_reada_buf(
 	struct xfs_trans	*trans,
 	struct xfs_inode	*dp,
 	xfs_dablk_t		bno,
-	int			whichfork)
+	xfs_daddr_t		mappedbno,
+	int			whichfork,
+	const struct xfs_buf_ops *ops)
 {
-	xfs_daddr_t		mappedbno = -1;
 	struct xfs_buf_map	map;
 	struct xfs_buf_map	*mapp;
 	int			nmap;
@@ -2221,7 +2322,7 @@ xfs_da_reada_buf(
 
 	mapp = ↦
 	nmap = 1;
-	error = xfs_dabuf_map(trans, dp, bno, -1, whichfork,
+	error = xfs_dabuf_map(trans, dp, bno, mappedbno, whichfork,
 				&mapp, &nmap);
 	if (error) {
 		/* mapping a hole is not an error, but we don't continue */
@@ -2231,7 +2332,7 @@ xfs_da_reada_buf(
 	}
 
 	mappedbno = mapp[0].bm_bn;
-	xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap);
+	xfs_buf_readahead_map(dp->i_mount->m_ddev_targp, mapp, nmap, ops);
 
 out_free:
 	if (mapp != &map)

fs/xfs/xfs_da_btree.h

@@ -18,7 +18,6 @@
 #ifndef __XFS_DA_BTREE_H__
 #define	__XFS_DA_BTREE_H__
 
-struct xfs_buf;
 struct xfs_bmap_free;
 struct xfs_inode;
 struct xfs_mount;
@@ -214,6 +213,9 @@ int	xfs_da_path_shift(xfs_da_state_t *state, xfs_da_state_path_t *path,
  */
 int	xfs_da_blk_link(xfs_da_state_t *state, xfs_da_state_blk_t *old_blk,
 				       xfs_da_state_blk_t *new_blk);
+int	xfs_da_node_read(struct xfs_trans *tp, struct xfs_inode *dp,
+			 xfs_dablk_t bno, xfs_daddr_t mappedbno,
+			 struct xfs_buf **bpp, int which_fork);
 
 /*
  * Utility routines.
@@ -226,9 +228,11 @@ int	xfs_da_get_buf(struct xfs_trans *trans, struct xfs_inode *dp,
 			      struct xfs_buf **bp, int whichfork);
 int	xfs_da_read_buf(struct xfs_trans *trans, struct xfs_inode *dp,
 			       xfs_dablk_t bno, xfs_daddr_t mappedbno,
-			       struct xfs_buf **bpp, int whichfork);
+			       struct xfs_buf **bpp, int whichfork,
+			       const struct xfs_buf_ops *ops);
 xfs_daddr_t	xfs_da_reada_buf(struct xfs_trans *trans, struct xfs_inode *dp,
-			xfs_dablk_t bno, int whichfork);
+				xfs_dablk_t bno, xfs_daddr_t mapped_bno,
+				int whichfork, const struct xfs_buf_ops *ops);
 int	xfs_da_shrink_inode(xfs_da_args_t *args, xfs_dablk_t dead_blkno,
 					  struct xfs_buf *dead_buf);
 

fs/xfs/xfs_dfrag.c

@@ -246,12 +246,10 @@ xfs_swap_extents(
 		goto out_unlock;
 	}
 
-	if (VN_CACHED(VFS_I(tip)) != 0) {
-		error = xfs_flushinval_pages(tip, 0, -1,
-				FI_REMAPF_LOCKED);
-		if (error)
-			goto out_unlock;
-	}
+	error = -filemap_write_and_wait(VFS_I(ip)->i_mapping);
+	if (error)
+		goto out_unlock;
+	truncate_pagecache_range(VFS_I(ip), 0, -1);
 
 	/* Verify O_DIRECT for ftmp */
 	if (VN_CACHED(VFS_I(tip)) != 0) {
@@ -315,8 +313,7 @@ xfs_swap_extents(
 	 * are safe.  We don't really care if non-io related
 	 * fields change.
 	 */
-
-	xfs_tosspages(ip, 0, -1, FI_REMAPF);
+	truncate_pagecache_range(VFS_I(ip), 0, -1);
 
 	tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
 	if ((error = xfs_trans_reserve(tp, 0,

fs/xfs/xfs_dir2_block.c

@@ -56,6 +56,214 @@ xfs_dir_startup(void)
 	xfs_dir_hash_dotdot = xfs_da_hashname((unsigned char *)"..", 2);
 }
 
+static void
+xfs_dir2_block_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_dir2_data_hdr *hdr = bp->b_addr;
+	int			block_ok = 0;
+
+	block_ok = hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
+	block_ok = block_ok && __xfs_dir2_data_check(NULL, bp) == 0;
+
+	if (!block_ok) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_dir2_block_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_block_verify(bp);
+}
+
+static void
+xfs_dir2_block_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_block_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_dir2_block_buf_ops = {
+	.verify_read = xfs_dir2_block_read_verify,
+	.verify_write = xfs_dir2_block_write_verify,
+};
+
+static int
+xfs_dir2_block_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	struct xfs_buf		**bpp)
+{
+	struct xfs_mount	*mp = dp->i_mount;
+
+	return xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, bpp,
+				XFS_DATA_FORK, &xfs_dir2_block_buf_ops);
+}
+
+static void
+xfs_dir2_block_need_space(
+	struct xfs_dir2_data_hdr	*hdr,
+	struct xfs_dir2_block_tail	*btp,
+	struct xfs_dir2_leaf_entry	*blp,
+	__be16				**tagpp,
+	struct xfs_dir2_data_unused	**dupp,
+	struct xfs_dir2_data_unused	**enddupp,
+	int				*compact,
+	int				len)
+{
+	struct xfs_dir2_data_free	*bf;
+	__be16				*tagp = NULL;
+	struct xfs_dir2_data_unused	*dup = NULL;
+	struct xfs_dir2_data_unused	*enddup = NULL;
+
+	*compact = 0;
+	bf = hdr->bestfree;
+
+	/*
+	 * If there are stale entries we'll use one for the leaf.
+	 */
+	if (btp->stale) {
+		if (be16_to_cpu(bf[0].length) >= len) {
+			/*
+			 * The biggest entry enough to avoid compaction.
+			 */
+			dup = (xfs_dir2_data_unused_t *)
+			      ((char *)hdr + be16_to_cpu(bf[0].offset));
+			goto out;
+		}
+
+		/*
+		 * Will need to compact to make this work.
+		 * Tag just before the first leaf entry.
+		 */
+		*compact = 1;
+		tagp = (__be16 *)blp - 1;
+
+		/* Data object just before the first leaf entry.  */
+		dup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
+
+		/*
+		 * If it's not free then the data will go where the
+		 * leaf data starts now, if it works at all.
+		 */
+		if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
+			if (be16_to_cpu(dup->length) + (be32_to_cpu(btp->stale) - 1) *
+			    (uint)sizeof(*blp) < len)
+				dup = NULL;
+		} else if ((be32_to_cpu(btp->stale) - 1) * (uint)sizeof(*blp) < len)
+			dup = NULL;
+		else
+			dup = (xfs_dir2_data_unused_t *)blp;
+		goto out;
+	}
+
+	/*
+	 * no stale entries, so just use free space.
+	 * Tag just before the first leaf entry.
+	 */
+	tagp = (__be16 *)blp - 1;
+
+	/* Data object just before the first leaf entry.  */
+	enddup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
+
+	/*
+	 * If it's not free then can't do this add without cleaning up:
+	 * the space before the first leaf entry needs to be free so it
+	 * can be expanded to hold the pointer to the new entry.
+	 */
+	if (be16_to_cpu(enddup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
+		/*
+		 * Check out the biggest freespace and see if it's the same one.
+		 */
+		dup = (xfs_dir2_data_unused_t *)
+		      ((char *)hdr + be16_to_cpu(bf[0].offset));
+		if (dup != enddup) {
+			/*
+			 * Not the same free entry, just check its length.
+			 */
+			if (be16_to_cpu(dup->length) < len)
+				dup = NULL;
+			goto out;
+		}
+
+		/*
+		 * It is the biggest freespace, can it hold the leaf too?
+		 */
+		if (be16_to_cpu(dup->length) < len + (uint)sizeof(*blp)) {
+			/*
+			 * Yes, use the second-largest entry instead if it works.
+			 */
+			if (be16_to_cpu(bf[1].length) >= len)
+				dup = (xfs_dir2_data_unused_t *)
+				      ((char *)hdr + be16_to_cpu(bf[1].offset));
+			else
+				dup = NULL;
+		}
+	}
+out:
+	*tagpp = tagp;
+	*dupp = dup;
+	*enddupp = enddup;
+}
+
+/*
+ * compact the leaf entries.
+ * Leave the highest-numbered stale entry stale.
+ * XXX should be the one closest to mid but mid is not yet computed.
+ */
+static void
+xfs_dir2_block_compact(
+	struct xfs_trans		*tp,
+	struct xfs_buf			*bp,
+	struct xfs_dir2_data_hdr	*hdr,
+	struct xfs_dir2_block_tail	*btp,
+	struct xfs_dir2_leaf_entry	*blp,
+	int				*needlog,
+	int				*lfloghigh,
+	int				*lfloglow)
+{
+	int			fromidx;	/* source leaf index */
+	int			toidx;		/* target leaf index */
+	int			needscan = 0;
+	int			highstale;	/* high stale index */
+
+	fromidx = toidx = be32_to_cpu(btp->count) - 1;
+	highstale = *lfloghigh = -1;
+	for (; fromidx >= 0; fromidx--) {
+		if (blp[fromidx].address == cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
+			if (highstale == -1)
+				highstale = toidx;
+			else {
+				if (*lfloghigh == -1)
+					*lfloghigh = toidx;
+				continue;
+			}
+		}
+		if (fromidx < toidx)
+			blp[toidx] = blp[fromidx];
+		toidx--;
+	}
+	*lfloglow = toidx + 1 - (be32_to_cpu(btp->stale) - 1);
+	*lfloghigh -= be32_to_cpu(btp->stale) - 1;
+	be32_add_cpu(&btp->count, -(be32_to_cpu(btp->stale) - 1));
+	xfs_dir2_data_make_free(tp, bp,
+		(xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
+		(xfs_dir2_data_aoff_t)((be32_to_cpu(btp->stale) - 1) * sizeof(*blp)),
+		needlog, &needscan);
+	blp += be32_to_cpu(btp->stale) - 1;
+	btp->stale = cpu_to_be32(1);
+	/*
+	 * If we now need to rebuild the bestfree map, do so.
+	 * This needs to happen before the next call to use_free.
+	 */
+	if (needscan)
+		xfs_dir2_data_freescan(tp->t_mountp, hdr, needlog);
+}
+
 /*
  * Add an entry to a block directory.
  */
@@ -63,7 +271,6 @@ int						/* error */
 xfs_dir2_block_addname(
 	xfs_da_args_t		*args)		/* directory op arguments */
 {
-	xfs_dir2_data_free_t	*bf;		/* bestfree table in block */
 	xfs_dir2_data_hdr_t	*hdr;		/* block header */
 	xfs_dir2_leaf_entry_t	*blp;		/* block leaf entries */
 	struct xfs_buf		*bp;		/* buffer for block */
@@ -94,134 +301,44 @@ xfs_dir2_block_addname(
 	dp = args->dp;
 	tp = args->trans;
 	mp = dp->i_mount;
-	/*
-	 * Read the (one and only) directory block into dabuf bp.
-	 */
-	if ((error =
-	    xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp, XFS_DATA_FORK))) {
+
+	/* Read the (one and only) directory block into bp. */
+	error = xfs_dir2_block_read(tp, dp, &bp);
+	if (error)
 		return error;
-	}
-	ASSERT(bp != NULL);
-	hdr = bp->b_addr;
-	/*
-	 * Check the magic number, corrupted if wrong.
-	 */
-	if (unlikely(hdr->magic != cpu_to_be32(XFS_DIR2_BLOCK_MAGIC))) {
-		XFS_CORRUPTION_ERROR("xfs_dir2_block_addname",
-				     XFS_ERRLEVEL_LOW, mp, hdr);
-		xfs_trans_brelse(tp, bp);
-		return XFS_ERROR(EFSCORRUPTED);
-	}
+
 	len = xfs_dir2_data_entsize(args->namelen);
+
 	/*
 	 * Set up pointers to parts of the block.
 	 */
-	bf = hdr->bestfree;
+	hdr = bp->b_addr;
 	btp = xfs_dir2_block_tail_p(mp, hdr);
 	blp = xfs_dir2_block_leaf_p(btp);
+
 	/*
-	 * No stale entries?  Need space for entry and new leaf.
-	 */
-	if (!btp->stale) {
-		/*
-		 * Tag just before the first leaf entry.
-		 */
-		tagp = (__be16 *)blp - 1;
-		/*
-		 * Data object just before the first leaf entry.
-		 */
-		enddup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
-		/*
-		 * If it's not free then can't do this add without cleaning up:
-		 * the space before the first leaf entry needs to be free so it
-		 * can be expanded to hold the pointer to the new entry.
-		 */
-		if (be16_to_cpu(enddup->freetag) != XFS_DIR2_DATA_FREE_TAG)
-			dup = enddup = NULL;
-		/*
-		 * Check out the biggest freespace and see if it's the same one.
-		 */
-		else {
-			dup = (xfs_dir2_data_unused_t *)
-			      ((char *)hdr + be16_to_cpu(bf[0].offset));
-			if (dup == enddup) {
-				/*
-				 * It is the biggest freespace, is it too small
-				 * to hold the new leaf too?
-				 */
-				if (be16_to_cpu(dup->length) < len + (uint)sizeof(*blp)) {
-					/*
-					 * Yes, we use the second-largest
-					 * entry instead if it works.
-					 */
-					if (be16_to_cpu(bf[1].length) >= len)
-						dup = (xfs_dir2_data_unused_t *)
-						      ((char *)hdr +
-						       be16_to_cpu(bf[1].offset));
-					else
-						dup = NULL;
-				}
-			} else {
-				/*
-				 * Not the same free entry,
-				 * just check its length.
-				 */
-				if (be16_to_cpu(dup->length) < len) {
-					dup = NULL;
-				}
-			}
-		}
-		compact = 0;
-	}
-	/*
-	 * If there are stale entries we'll use one for the leaf.
-	 * Is the biggest entry enough to avoid compaction?
-	 */
-	else if (be16_to_cpu(bf[0].length) >= len) {
-		dup = (xfs_dir2_data_unused_t *)
-		      ((char *)hdr + be16_to_cpu(bf[0].offset));
-		compact = 0;
-	}
-	/*
-	 * Will need to compact to make this work.
+	 * Find out if we can reuse stale entries or whether we need extra
+	 * space for entry and new leaf.
 	 */
-	else {
-		/*
-		 * Tag just before the first leaf entry.
-		 */
-		tagp = (__be16 *)blp - 1;
-		/*
-		 * Data object just before the first leaf entry.
-		 */
-		dup = (xfs_dir2_data_unused_t *)((char *)hdr + be16_to_cpu(*tagp));
-		/*
-		 * If it's not free then the data will go where the
-		 * leaf data starts now, if it works at all.
-		 */
-		if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
-			if (be16_to_cpu(dup->length) + (be32_to_cpu(btp->stale) - 1) *
-			    (uint)sizeof(*blp) < len)
-				dup = NULL;
-		} else if ((be32_to_cpu(btp->stale) - 1) * (uint)sizeof(*blp) < len)
-			dup = NULL;
-		else
-			dup = (xfs_dir2_data_unused_t *)blp;
-		compact = 1;
-	}
+	xfs_dir2_block_need_space(hdr, btp, blp, &tagp, &dup,
+				  &enddup, &compact, len);
+
 	/*
-	 * If this isn't a real add, we're done with the buffer.
+	 * Done everything we need for a space check now.
 	 */
-	if (args->op_flags & XFS_DA_OP_JUSTCHECK)
+	if (args->op_flags & XFS_DA_OP_JUSTCHECK) {
 		xfs_trans_brelse(tp, bp);
+		if (!dup)
+			return XFS_ERROR(ENOSPC);
+		return 0;
+	}
+
 	/*
 	 * If we don't have space for the new entry & leaf ...
 	 */
 	if (!dup) {
-		/*
-		 * Not trying to actually do anything, or don't have
-		 * a space reservation: return no-space.
-		 */
-		if ((args->op_flags & XFS_DA_OP_JUSTCHECK) || args->total == 0)
+		/* Don't have a space reservation: return no-space.  */
+		if (args->total == 0)
 			return XFS_ERROR(ENOSPC);
 		/*
 		 * Convert to the next larger format.
@@ -232,65 +349,24 @@ xfs_dir2_block_addname(
 			return error;
 		return xfs_dir2_leaf_addname(args);
 	}
-	/*
-	 * Just checking, and it would work, so say so.
-	 */
-	if (args->op_flags & XFS_DA_OP_JUSTCHECK)
-		return 0;
+
 	needlog = needscan = 0;
+
 	/*
 	 * If need to compact the leaf entries, do it now.
-	 * Leave the highest-numbered stale entry stale.
-	 * XXX should be the one closest to mid but mid is not yet computed.
-	 */
-	if (compact) {
-		int	fromidx;		/* source leaf index */
-		int	toidx;			/* target leaf index */
-
-		for (fromidx = toidx = be32_to_cpu(btp->count) - 1,
-			highstale = lfloghigh = -1;
-		     fromidx >= 0;
-		     fromidx--) {
-			if (blp[fromidx].address ==
-			    cpu_to_be32(XFS_DIR2_NULL_DATAPTR)) {
-				if (highstale == -1)
-					highstale = toidx;
-				else {
-					if (lfloghigh == -1)
-						lfloghigh = toidx;
-					continue;
-				}
-			}
-			if (fromidx < toidx)
-				blp[toidx] = blp[fromidx];
-			toidx--;
-		}
-		lfloglow = toidx + 1 - (be32_to_cpu(btp->stale) - 1);
-		lfloghigh -= be32_to_cpu(btp->stale) - 1;
-		be32_add_cpu(&btp->count, -(be32_to_cpu(btp->stale) - 1));
-		xfs_dir2_data_make_free(tp, bp,
-			(xfs_dir2_data_aoff_t)((char *)blp - (char *)hdr),
-			(xfs_dir2_data_aoff_t)((be32_to_cpu(btp->stale) - 1) * sizeof(*blp)),
-			&needlog, &needscan);
-		blp += be32_to_cpu(btp->stale) - 1;
-		btp->stale = cpu_to_be32(1);
-		/*
-		 * If we now need to rebuild the bestfree map, do so.
-		 * This needs to happen before the next call to use_free.
-		 */
-		if (needscan) {
-			xfs_dir2_data_freescan(mp, hdr, &needlog);
-			needscan = 0;
-		}
-	}
-	/*
-	 * Set leaf logging boundaries to impossible state.
-	 * For the no-stale case they're set explicitly.
 	 */
+	if (compact)
+		xfs_dir2_block_compact(tp, bp, hdr, btp, blp, &needlog,
+				      &lfloghigh, &lfloglow);
 	else if (btp->stale) {
+		/*
+		 * Set leaf logging boundaries to impossible state.
+		 * For the no-stale case they're set explicitly.
+		 */
 		lfloglow = be32_to_cpu(btp->count);
 		lfloghigh = -1;
 	}
+
 	/*
 	 * Find the slot that's first lower than our hash value, -1 if none.
 	 */
@@ -450,18 +526,13 @@ xfs_dir2_block_getdents(
 	/*
 	 * If the block number in the offset is out of range, we're done.
 	 */
-	if (xfs_dir2_dataptr_to_db(mp, *offset) > mp->m_dirdatablk) {
+	if (xfs_dir2_dataptr_to_db(mp, *offset) > mp->m_dirdatablk)
 		return 0;
-	}
-	/*
-	 * Can't read the block, give up, else get dabuf in bp.
-	 */
-	error = xfs_da_read_buf(NULL, dp, mp->m_dirdatablk, -1,
-				&bp, XFS_DATA_FORK);
+
+	error = xfs_dir2_block_read(NULL, dp, &bp);
 	if (error)
 		return error;
 
-	ASSERT(bp != NULL);
 	/*
 	 * Extract the byte offset we start at from the seek pointer.
 	 * We'll skip entries before this.
@@ -637,14 +708,11 @@ xfs_dir2_block_lookup_int(
 	dp = args->dp;
 	tp = args->trans;
 	mp = dp->i_mount;
-	/*
-	 * Read the buffer, return error if we can't get it.
-	 */
-	if ((error =
-	    xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &bp, XFS_DATA_FORK))) {
+
+	error = xfs_dir2_block_read(tp, dp, &bp);
+	if (error)
 		return error;
-	}
-	ASSERT(bp != NULL);
+
 	hdr = bp->b_addr;
 	xfs_dir2_data_check(dp, bp);
 	btp = xfs_dir2_block_tail_p(mp, hdr);
@@ -917,10 +985,10 @@ xfs_dir2_leaf_to_block(
 	/*
 	 * Read the data block if we don't already have it, give up if it fails.
 	 */
-	if (dbp == NULL &&
-	    (error = xfs_da_read_buf(tp, dp, mp->m_dirdatablk, -1, &dbp,
-		    XFS_DATA_FORK))) {
-		return error;
+	if (!dbp) {
+		error = xfs_dir2_data_read(tp, dp, mp->m_dirdatablk, -1, &dbp);
+		if (error)
+			return error;
 	}
 	hdr = dbp->b_addr;
 	ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC));
@@ -944,6 +1012,7 @@ xfs_dir2_leaf_to_block(
 	/*
 	 * Start converting it to block form.
 	 */
+	dbp->b_ops = &xfs_dir2_block_buf_ops;
 	hdr->magic = cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
 	needlog = 1;
 	needscan = 0;
@@ -1073,6 +1142,7 @@ xfs_dir2_sf_to_block(
 		kmem_free(sfp);
 		return error;
 	}
+	bp->b_ops = &xfs_dir2_block_buf_ops;
 	hdr = bp->b_addr;
 	hdr->magic = cpu_to_be32(XFS_DIR2_BLOCK_MAGIC);
 	/*

fs/xfs/xfs_dir2_data.c

@@ -34,14 +34,13 @@
 STATIC xfs_dir2_data_free_t *
 xfs_dir2_data_freefind(xfs_dir2_data_hdr_t *hdr, xfs_dir2_data_unused_t *dup);
 
-#ifdef DEBUG
 /*
  * Check the consistency of the data block.
  * The input can also be a block-format directory.
- * Pop an assert if we find anything bad.
+ * Return 0 is the buffer is good, otherwise an error.
  */
-void
-xfs_dir2_data_check(
+int
+__xfs_dir2_data_check(
 	struct xfs_inode	*dp,		/* incore inode pointer */
 	struct xfs_buf		*bp)		/* data block's buffer */
 {
@@ -64,18 +63,23 @@ xfs_dir2_data_check(
 	int			stale;		/* count of stale leaves */
 	struct xfs_name		name;
 
-	mp = dp->i_mount;
+	mp = bp->b_target->bt_mount;
 	hdr = bp->b_addr;
 	bf = hdr->bestfree;
 	p = (char *)(hdr + 1);
 
-	if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC)) {
+	switch (hdr->magic) {
+	case cpu_to_be32(XFS_DIR2_BLOCK_MAGIC):
 		btp = xfs_dir2_block_tail_p(mp, hdr);
 		lep = xfs_dir2_block_leaf_p(btp);
 		endp = (char *)lep;
-	} else {
-		ASSERT(hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC));
+		break;
+	case cpu_to_be32(XFS_DIR2_DATA_MAGIC):
 		endp = (char *)hdr + mp->m_dirblksize;
+		break;
+	default:
+		XFS_ERROR_REPORT("Bad Magic", XFS_ERRLEVEL_LOW, mp);
+		return EFSCORRUPTED;
 	}
 
 	count = lastfree = freeseen = 0;
@@ -83,19 +87,22 @@ xfs_dir2_data_check(
 	 * Account for zero bestfree entries.
 	 */
 	if (!bf[0].length) {
-		ASSERT(!bf[0].offset);
+		XFS_WANT_CORRUPTED_RETURN(!bf[0].offset);
 		freeseen |= 1 << 0;
 	}
 	if (!bf[1].length) {
-		ASSERT(!bf[1].offset);
+		XFS_WANT_CORRUPTED_RETURN(!bf[1].offset);
 		freeseen |= 1 << 1;
 	}
 	if (!bf[2].length) {
-		ASSERT(!bf[2].offset);
+		XFS_WANT_CORRUPTED_RETURN(!bf[2].offset);
 		freeseen |= 1 << 2;
 	}
-	ASSERT(be16_to_cpu(bf[0].length) >= be16_to_cpu(bf[1].length));
-	ASSERT(be16_to_cpu(bf[1].length) >= be16_to_cpu(bf[2].length));
+
+	XFS_WANT_CORRUPTED_RETURN(be16_to_cpu(bf[0].length) >=
+						be16_to_cpu(bf[1].length));
+	XFS_WANT_CORRUPTED_RETURN(be16_to_cpu(bf[1].length) >=
+						be16_to_cpu(bf[2].length));
 	/*
 	 * Loop over the data/unused entries.
 	 */
@@ -107,17 +114,20 @@ xfs_dir2_data_check(
 		 * doesn't need to be there.
 		 */
 		if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) {
-			ASSERT(lastfree == 0);
-			ASSERT(be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) ==
-			       (char *)dup - (char *)hdr);
+			XFS_WANT_CORRUPTED_RETURN(lastfree == 0);
+			XFS_WANT_CORRUPTED_RETURN(
+				be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) ==
+					       (char *)dup - (char *)hdr);
 			dfp = xfs_dir2_data_freefind(hdr, dup);
 			if (dfp) {
 				i = (int)(dfp - bf);
-				ASSERT((freeseen & (1 << i)) == 0);
+				XFS_WANT_CORRUPTED_RETURN(
+					(freeseen & (1 << i)) == 0);
 				freeseen |= 1 << i;
 			} else {
-				ASSERT(be16_to_cpu(dup->length) <=
-				       be16_to_cpu(bf[2].length));
+				XFS_WANT_CORRUPTED_RETURN(
+					be16_to_cpu(dup->length) <=
+						be16_to_cpu(bf[2].length));
 			}
 			p += be16_to_cpu(dup->length);
 			lastfree = 1;
@@ -130,10 +140,12 @@ xfs_dir2_data_check(
 		 * The linear search is crude but this is DEBUG code.
 		 */
 		dep = (xfs_dir2_data_entry_t *)p;
-		ASSERT(dep->namelen != 0);
-		ASSERT(xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber)) == 0);
-		ASSERT(be16_to_cpu(*xfs_dir2_data_entry_tag_p(dep)) ==
-		       (char *)dep - (char *)hdr);
+		XFS_WANT_CORRUPTED_RETURN(dep->namelen != 0);
+		XFS_WANT_CORRUPTED_RETURN(
+			!xfs_dir_ino_validate(mp, be64_to_cpu(dep->inumber)));
+		XFS_WANT_CORRUPTED_RETURN(
+			be16_to_cpu(*xfs_dir2_data_entry_tag_p(dep)) ==
+					       (char *)dep - (char *)hdr);
 		count++;
 		lastfree = 0;
 		if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC)) {
@@ -148,27 +160,122 @@ xfs_dir2_data_check(
 				    be32_to_cpu(lep[i].hashval) == hash)
 					break;
 			}
-			ASSERT(i < be32_to_cpu(btp->count));
+			XFS_WANT_CORRUPTED_RETURN(i < be32_to_cpu(btp->count));
 		}
 		p += xfs_dir2_data_entsize(dep->namelen);
 	}
 	/*
 	 * Need to have seen all the entries and all the bestfree slots.
 	 */
-	ASSERT(freeseen == 7);
+	XFS_WANT_CORRUPTED_RETURN(freeseen == 7);
 	if (hdr->magic == cpu_to_be32(XFS_DIR2_BLOCK_MAGIC)) {
 		for (i = stale = 0; i < be32_to_cpu(btp->count); i++) {
 			if (lep[i].address ==
 			    cpu_to_be32(XFS_DIR2_NULL_DATAPTR))
 				stale++;
 			if (i > 0)
-				ASSERT(be32_to_cpu(lep[i].hashval) >= be32_to_cpu(lep[i - 1].hashval));
+				XFS_WANT_CORRUPTED_RETURN(
+					be32_to_cpu(lep[i].hashval) >=
+						be32_to_cpu(lep[i - 1].hashval));
 		}
-		ASSERT(count == be32_to_cpu(btp->count) - be32_to_cpu(btp->stale));
-		ASSERT(stale == be32_to_cpu(btp->stale));
+		XFS_WANT_CORRUPTED_RETURN(count ==
+			be32_to_cpu(btp->count) - be32_to_cpu(btp->stale));
+		XFS_WANT_CORRUPTED_RETURN(stale == be32_to_cpu(btp->stale));
 	}
+	return 0;
+}
+
+static void
+xfs_dir2_data_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_dir2_data_hdr *hdr = bp->b_addr;
+	int			block_ok = 0;
+
+	block_ok = hdr->magic == cpu_to_be32(XFS_DIR2_DATA_MAGIC);
+	block_ok = block_ok && __xfs_dir2_data_check(NULL, bp) == 0;
+
+	if (!block_ok) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+/*
+ * Readahead of the first block of the directory when it is opened is completely
+ * oblivious to the format of the directory. Hence we can either get a block
+ * format buffer or a data format buffer on readahead.
+ */
+static void
+xfs_dir2_data_reada_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_dir2_data_hdr *hdr = bp->b_addr;
+
+	switch (hdr->magic) {
+	case cpu_to_be32(XFS_DIR2_BLOCK_MAGIC):
+		bp->b_ops = &xfs_dir2_block_buf_ops;
+		bp->b_ops->verify_read(bp);
+		return;
+	case cpu_to_be32(XFS_DIR2_DATA_MAGIC):
+		xfs_dir2_data_verify(bp);
+		return;
+	default:
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+		break;
+	}
+}
+
+static void
+xfs_dir2_data_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_data_verify(bp);
+}
+
+static void
+xfs_dir2_data_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_data_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_dir2_data_buf_ops = {
+	.verify_read = xfs_dir2_data_read_verify,
+	.verify_write = xfs_dir2_data_write_verify,
+};
+
+static const struct xfs_buf_ops xfs_dir2_data_reada_buf_ops = {
+	.verify_read = xfs_dir2_data_reada_verify,
+	.verify_write = xfs_dir2_data_write_verify,
+};
+
+
+int
+xfs_dir2_data_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		bno,
+	xfs_daddr_t		mapped_bno,
+	struct xfs_buf		**bpp)
+{
+	return xfs_da_read_buf(tp, dp, bno, mapped_bno, bpp,
+				XFS_DATA_FORK, &xfs_dir2_data_buf_ops);
+}
+
+int
+xfs_dir2_data_readahead(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		bno,
+	xfs_daddr_t		mapped_bno)
+{
+	return xfs_da_reada_buf(tp, dp, bno, mapped_bno,
+				XFS_DATA_FORK, &xfs_dir2_data_reada_buf_ops);
 }
-#endif
 
 /*
  * Given a data block and an unused entry from that block,
@@ -409,10 +516,9 @@ xfs_dir2_data_init(
 	 */
 	error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, blkno), -1, &bp,
 		XFS_DATA_FORK);
-	if (error) {
+	if (error)
 		return error;
-	}
-	ASSERT(bp != NULL);
+	bp->b_ops = &xfs_dir2_data_buf_ops;
 
 	/*
 	 * Initialize the header.

fs/xfs/xfs_dir2_leaf.c

@@ -48,6 +48,83 @@ static void xfs_dir2_leaf_log_bests(struct xfs_trans *tp, struct xfs_buf *bp,
 				    int first, int last);
 static void xfs_dir2_leaf_log_tail(struct xfs_trans *tp, struct xfs_buf *bp);
 
+static void
+xfs_dir2_leaf_verify(
+	struct xfs_buf		*bp,
+	__be16			magic)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_dir2_leaf_hdr *hdr = bp->b_addr;
+	int			block_ok = 0;
+
+	block_ok = hdr->info.magic == magic;
+	if (!block_ok) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_dir2_leaf1_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
+}
+
+static void
+xfs_dir2_leaf1_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAF1_MAGIC));
+}
+
+void
+xfs_dir2_leafn_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
+}
+
+void
+xfs_dir2_leafn_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_leaf_verify(bp, cpu_to_be16(XFS_DIR2_LEAFN_MAGIC));
+}
+
+static const struct xfs_buf_ops xfs_dir2_leaf1_buf_ops = {
+	.verify_read = xfs_dir2_leaf1_read_verify,
+	.verify_write = xfs_dir2_leaf1_write_verify,
+};
+
+const struct xfs_buf_ops xfs_dir2_leafn_buf_ops = {
+	.verify_read = xfs_dir2_leafn_read_verify,
+	.verify_write = xfs_dir2_leafn_write_verify,
+};
+
+static int
+xfs_dir2_leaf_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		fbno,
+	xfs_daddr_t		mappedbno,
+	struct xfs_buf		**bpp)
+{
+	return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
+				XFS_DATA_FORK, &xfs_dir2_leaf1_buf_ops);
+}
+
+int
+xfs_dir2_leafn_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		fbno,
+	xfs_daddr_t		mappedbno,
+	struct xfs_buf		**bpp)
+{
+	return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
+				XFS_DATA_FORK, &xfs_dir2_leafn_buf_ops);
+}
 
 /*
  * Convert a block form directory to a leaf form directory.
@@ -125,6 +202,7 @@ xfs_dir2_block_to_leaf(
 	/*
 	 * Fix up the block header, make it a data block.
 	 */
+	dbp->b_ops = &xfs_dir2_data_buf_ops;
 	hdr->magic = cpu_to_be32(XFS_DIR2_DATA_MAGIC);
 	if (needscan)
 		xfs_dir2_data_freescan(mp, hdr, &needlog);
@@ -311,15 +389,11 @@ xfs_dir2_leaf_addname(
 	dp = args->dp;
 	tp = args->trans;
 	mp = dp->i_mount;
-	/*
-	 * Read the leaf block.
-	 */
-	error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
-		XFS_DATA_FORK);
-	if (error) {
+
+	error = xfs_dir2_leaf_read(tp, dp, mp->m_dirleafblk, -1, &lbp);
+	if (error)
 		return error;
-	}
-	ASSERT(lbp != NULL);
+
 	/*
 	 * Look up the entry by hash value and name.
 	 * We know it's not there, our caller has already done a lookup.
@@ -494,22 +568,21 @@ xfs_dir2_leaf_addname(
 		hdr = dbp->b_addr;
 		bestsp[use_block] = hdr->bestfree[0].length;
 		grown = 1;
-	}
-	/*
-	 * Already had space in some data block.
-	 * Just read that one in.
-	 */
-	else {
-		if ((error =
-		    xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, use_block),
-			    -1, &dbp, XFS_DATA_FORK))) {
+	} else {
+		/*
+		 * Already had space in some data block.
+		 * Just read that one in.
+		 */
+		error = xfs_dir2_data_read(tp, dp,
+					   xfs_dir2_db_to_da(mp, use_block),
+					   -1, &dbp);
+		if (error) {
 			xfs_trans_brelse(tp, lbp);
 			return error;
 		}
 		hdr = dbp->b_addr;
 		grown = 0;
 	}
-	xfs_dir2_data_check(dp, dbp);
 	/*
 	 * Point to the biggest freespace in our data block.
 	 */
@@ -892,10 +965,9 @@ xfs_dir2_leaf_readbuf(
 	 * Read the directory block starting at the first mapping.
 	 */
 	mip->curdb = xfs_dir2_da_to_db(mp, map->br_startoff);
-	error = xfs_da_read_buf(NULL, dp, map->br_startoff,
+	error = xfs_dir2_data_read(NULL, dp, map->br_startoff,
 			map->br_blockcount >= mp->m_dirblkfsbs ?
-			    XFS_FSB_TO_DADDR(mp, map->br_startblock) : -1,
-			&bp, XFS_DATA_FORK);
+			    XFS_FSB_TO_DADDR(mp, map->br_startblock) : -1, &bp);
 
 	/*
 	 * Should just skip over the data block instead of giving up.
@@ -922,11 +994,11 @@ xfs_dir2_leaf_readbuf(
 		 */
 		if (i > mip->ra_current &&
 		    map[mip->ra_index].br_blockcount >= mp->m_dirblkfsbs) {
-			xfs_buf_readahead(mp->m_ddev_targp,
+			xfs_dir2_data_readahead(NULL, dp,
+				map[mip->ra_index].br_startoff + mip->ra_offset,
 				XFS_FSB_TO_DADDR(mp,
 					map[mip->ra_index].br_startblock +
-							mip->ra_offset),
-				(int)BTOBB(mp->m_dirblksize));
+							mip->ra_offset));
 			mip->ra_current = i;
 		}
 
@@ -935,10 +1007,9 @@ xfs_dir2_leaf_readbuf(
 		 * use our mapping, but this is a very rare case.
 		 */
 		else if (i > mip->ra_current) {
-			xfs_da_reada_buf(NULL, dp,
+			xfs_dir2_data_readahead(NULL, dp,
 					map[mip->ra_index].br_startoff +
-							mip->ra_offset,
-					XFS_DATA_FORK);
+							mip->ra_offset, -1);
 			mip->ra_current = i;
 		}
 
@@ -1177,15 +1248,14 @@ xfs_dir2_leaf_init(
 	 * Get the buffer for the block.
 	 */
 	error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, bno), -1, &bp,
-		XFS_DATA_FORK);
-	if (error) {
+			       XFS_DATA_FORK);
+	if (error)
 		return error;
-	}
-	ASSERT(bp != NULL);
-	leaf = bp->b_addr;
+
 	/*
 	 * Initialize the header.
 	 */
+	leaf = bp->b_addr;
 	leaf->hdr.info.magic = cpu_to_be16(magic);
 	leaf->hdr.info.forw = 0;
 	leaf->hdr.info.back = 0;
@@ -1198,10 +1268,12 @@ xfs_dir2_leaf_init(
 	 * the block.
 	 */
 	if (magic == XFS_DIR2_LEAF1_MAGIC) {
+		bp->b_ops = &xfs_dir2_leaf1_buf_ops;
 		ltp = xfs_dir2_leaf_tail_p(mp, leaf);
 		ltp->bestcount = 0;
 		xfs_dir2_leaf_log_tail(tp, bp);
-	}
+	} else
+		bp->b_ops = &xfs_dir2_leafn_buf_ops;
 	*bpp = bp;
 	return 0;
 }
@@ -1372,13 +1444,11 @@ xfs_dir2_leaf_lookup_int(
 	dp = args->dp;
 	tp = args->trans;
 	mp = dp->i_mount;
-	/*
-	 * Read the leaf block into the buffer.
-	 */
-	error = xfs_da_read_buf(tp, dp, mp->m_dirleafblk, -1, &lbp,
-							XFS_DATA_FORK);
+
+	error = xfs_dir2_leaf_read(tp, dp, mp->m_dirleafblk, -1, &lbp);
 	if (error)
 		return error;
+
 	*lbpp = lbp;
 	leaf = lbp->b_addr;
 	xfs_dir2_leaf_check(dp, lbp);
@@ -1409,14 +1479,13 @@ xfs_dir2_leaf_lookup_int(
 		if (newdb != curdb) {
 			if (dbp)
 				xfs_trans_brelse(tp, dbp);
-			error = xfs_da_read_buf(tp, dp,
-						xfs_dir2_db_to_da(mp, newdb),
-						-1, &dbp, XFS_DATA_FORK);
+			error = xfs_dir2_data_read(tp, dp,
+						   xfs_dir2_db_to_da(mp, newdb),
+						   -1, &dbp);
 			if (error) {
 				xfs_trans_brelse(tp, lbp);
 				return error;
 			}
-			xfs_dir2_data_check(dp, dbp);
 			curdb = newdb;
 		}
 		/*
@@ -1451,9 +1520,9 @@ xfs_dir2_leaf_lookup_int(
 		ASSERT(cidb != -1);
 		if (cidb != curdb) {
 			xfs_trans_brelse(tp, dbp);
-			error = xfs_da_read_buf(tp, dp,
-						xfs_dir2_db_to_da(mp, cidb),
-						-1, &dbp, XFS_DATA_FORK);
+			error = xfs_dir2_data_read(tp, dp,
+						   xfs_dir2_db_to_da(mp, cidb),
+						   -1, &dbp);
 			if (error) {
 				xfs_trans_brelse(tp, lbp);
 				return error;
@@ -1738,10 +1807,9 @@ xfs_dir2_leaf_trim_data(
 	/*
 	 * Read the offending data block.  We need its buffer.
 	 */
-	if ((error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp,
-			XFS_DATA_FORK))) {
+	error = xfs_dir2_data_read(tp, dp, xfs_dir2_db_to_da(mp, db), -1, &dbp);
+	if (error)
 		return error;
-	}
 
 	leaf = lbp->b_addr;
 	ltp = xfs_dir2_leaf_tail_p(mp, leaf);
@@ -1864,10 +1932,9 @@ xfs_dir2_node_to_leaf(
 	/*
 	 * Read the freespace block.
 	 */
-	if ((error = xfs_da_read_buf(tp, dp, mp->m_dirfreeblk, -1, &fbp,
-			XFS_DATA_FORK))) {
+	error = xfs_dir2_free_read(tp, dp,  mp->m_dirfreeblk, &fbp);
+	if (error)
 		return error;
-	}
 	free = fbp->b_addr;
 	ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
 	ASSERT(!free->hdr.firstdb);
@@ -1890,7 +1957,10 @@ xfs_dir2_node_to_leaf(
 		xfs_dir2_leaf_compact(args, lbp);
 	else
 		xfs_dir2_leaf_log_header(tp, lbp);
+
+	lbp->b_ops = &xfs_dir2_leaf1_buf_ops;
 	leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAF1_MAGIC);
+
 	/*
 	 * Set up the leaf tail from the freespace block.
 	 */

fs/xfs/xfs_dir2_node.c

@@ -55,6 +55,74 @@ static int xfs_dir2_leafn_remove(xfs_da_args_t *args, struct xfs_buf *bp,
 static int xfs_dir2_node_addname_int(xfs_da_args_t *args,
 				     xfs_da_state_blk_t *fblk);
 
+static void
+xfs_dir2_free_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_dir2_free_hdr *hdr = bp->b_addr;
+	int			block_ok = 0;
+
+	block_ok = hdr->magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC);
+	if (!block_ok) {
+		XFS_CORRUPTION_ERROR("xfs_dir2_free_verify magic",
+				     XFS_ERRLEVEL_LOW, mp, hdr);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_dir2_free_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_free_verify(bp);
+}
+
+static void
+xfs_dir2_free_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dir2_free_verify(bp);
+}
+
+static const struct xfs_buf_ops xfs_dir2_free_buf_ops = {
+	.verify_read = xfs_dir2_free_read_verify,
+	.verify_write = xfs_dir2_free_write_verify,
+};
+
+
+static int
+__xfs_dir2_free_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		fbno,
+	xfs_daddr_t		mappedbno,
+	struct xfs_buf		**bpp)
+{
+	return xfs_da_read_buf(tp, dp, fbno, mappedbno, bpp,
+				XFS_DATA_FORK, &xfs_dir2_free_buf_ops);
+}
+
+int
+xfs_dir2_free_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		fbno,
+	struct xfs_buf		**bpp)
+{
+	return __xfs_dir2_free_read(tp, dp, fbno, -1, bpp);
+}
+
+static int
+xfs_dir2_free_try_read(
+	struct xfs_trans	*tp,
+	struct xfs_inode	*dp,
+	xfs_dablk_t		fbno,
+	struct xfs_buf		**bpp)
+{
+	return __xfs_dir2_free_read(tp, dp, fbno, -2, bpp);
+}
+
 /*
  * Log entries from a freespace block.
  */
@@ -131,11 +199,12 @@ xfs_dir2_leaf_to_node(
 	/*
 	 * Get the buffer for the new freespace block.
 	 */
-	if ((error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb), -1, &fbp,
-			XFS_DATA_FORK))) {
+	error = xfs_da_get_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb), -1, &fbp,
+				XFS_DATA_FORK);
+	if (error)
 		return error;
-	}
-	ASSERT(fbp != NULL);
+	fbp->b_ops = &xfs_dir2_free_buf_ops;
+
 	free = fbp->b_addr;
 	leaf = lbp->b_addr;
 	ltp = xfs_dir2_leaf_tail_p(mp, leaf);
@@ -157,7 +226,10 @@ xfs_dir2_leaf_to_node(
 		*to = cpu_to_be16(off);
 	}
 	free->hdr.nused = cpu_to_be32(n);
+
+	lbp->b_ops = &xfs_dir2_leafn_buf_ops;
 	leaf->hdr.info.magic = cpu_to_be16(XFS_DIR2_LEAFN_MAGIC);
+
 	/*
 	 * Log everything.
 	 */
@@ -394,12 +466,10 @@ xfs_dir2_leafn_lookup_for_addname(
 				 */
 				if (curbp)
 					xfs_trans_brelse(tp, curbp);
-				/*
-				 * Read the free block.
-				 */
-				error = xfs_da_read_buf(tp, dp,
+
+				error = xfs_dir2_free_read(tp, dp,
 						xfs_dir2_db_to_da(mp, newfdb),
-						-1, &curbp, XFS_DATA_FORK);
+						&curbp);
 				if (error)
 					return error;
 				free = curbp->b_addr;
@@ -534,9 +604,9 @@ xfs_dir2_leafn_lookup_for_entry(
 				ASSERT(state->extravalid);
 				curbp = state->extrablk.bp;
 			} else {
-				error = xfs_da_read_buf(tp, dp,
+				error = xfs_dir2_data_read(tp, dp,
 						xfs_dir2_db_to_da(mp, newdb),
-						-1, &curbp, XFS_DATA_FORK);
+						-1, &curbp);
 				if (error)
 					return error;
 			}
@@ -568,6 +638,7 @@ xfs_dir2_leafn_lookup_for_entry(
 			state->extrablk.index = (int)((char *)dep -
 							(char *)curbp->b_addr);
 			state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
+			curbp->b_ops = &xfs_dir2_data_buf_ops;
 			if (cmp == XFS_CMP_EXACT)
 				return XFS_ERROR(EEXIST);
 		}
@@ -582,6 +653,7 @@ xfs_dir2_leafn_lookup_for_entry(
 			state->extrablk.index = -1;
 			state->extrablk.blkno = curdb;
 			state->extrablk.magic = XFS_DIR2_DATA_MAGIC;
+			curbp->b_ops = &xfs_dir2_data_buf_ops;
 		} else {
 			/* If the curbp is not the CI match block, drop it */
 			if (state->extrablk.bp != curbp)
@@ -825,6 +897,77 @@ xfs_dir2_leafn_rebalance(
 	}
 }
 
+static int
+xfs_dir2_data_block_free(
+	xfs_da_args_t		*args,
+	struct xfs_dir2_data_hdr *hdr,
+	struct xfs_dir2_free	*free,
+	xfs_dir2_db_t		fdb,
+	int			findex,
+	struct xfs_buf		*fbp,
+	int			longest)
+{
+	struct xfs_trans	*tp = args->trans;
+	int			logfree = 0;
+
+	if (!hdr) {
+		/* One less used entry in the free table.  */
+		be32_add_cpu(&free->hdr.nused, -1);
+		xfs_dir2_free_log_header(tp, fbp);
+
+		/*
+		 * If this was the last entry in the table, we can trim the
+		 * table size back.  There might be other entries at the end
+		 * referring to non-existent data blocks, get those too.
+		 */
+		if (findex == be32_to_cpu(free->hdr.nvalid) - 1) {
+			int	i;		/* free entry index */
+
+			for (i = findex - 1; i >= 0; i--) {
+				if (free->bests[i] != cpu_to_be16(NULLDATAOFF))
+					break;
+			}
+			free->hdr.nvalid = cpu_to_be32(i + 1);
+			logfree = 0;
+		} else {
+			/* Not the last entry, just punch it out.  */
+			free->bests[findex] = cpu_to_be16(NULLDATAOFF);
+			logfree = 1;
+		}
+		/*
+		 * If there are no useful entries left in the block,
+		 * get rid of the block if we can.
+		 */
+		if (!free->hdr.nused) {
+			int error;
+
+			error = xfs_dir2_shrink_inode(args, fdb, fbp);
+			if (error == 0) {
+				fbp = NULL;
+				logfree = 0;
+			} else if (error != ENOSPC || args->total != 0)
+				return error;
+			/*
+			 * It's possible to get ENOSPC if there is no
+			 * space reservation.  In this case some one
+			 * else will eventually get rid of this block.
+			 */
+		}
+	} else {
+		/*
+		 * Data block is not empty, just set the free entry to the new
+		 * value.
+		 */
+		free->bests[findex] = cpu_to_be16(longest);
+		logfree = 1;
+	}
+
+	/* Log the free entry that changed, unless we got rid of it.  */
+	if (logfree)
+		xfs_dir2_free_log_bests(tp, fbp, findex, findex);
+	return 0;
+}
+
 /*
  * Remove an entry from a node directory.
  * This removes the leaf entry and the data entry,
@@ -908,17 +1051,16 @@ xfs_dir2_leafn_remove(
 		xfs_dir2_db_t	fdb;		/* freeblock block number */
 		int		findex;		/* index in freeblock entries */
 		xfs_dir2_free_t	*free;		/* freeblock structure */
-		int		logfree;	/* need to log free entry */
 
 		/*
 		 * Convert the data block number to a free block,
 		 * read in the free block.
 		 */
 		fdb = xfs_dir2_db_to_fdb(mp, db);
-		if ((error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, fdb),
-				-1, &fbp, XFS_DATA_FORK))) {
+		error = xfs_dir2_free_read(tp, dp, xfs_dir2_db_to_da(mp, fdb),
+					   &fbp);
+		if (error)
 			return error;
-		}
 		free = fbp->b_addr;
 		ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
 		ASSERT(be32_to_cpu(free->hdr.firstdb) ==
@@ -954,68 +1096,12 @@ xfs_dir2_leafn_remove(
 		 * If we got rid of the data block, we can eliminate that entry
 		 * in the free block.
 		 */
-		if (hdr == NULL) {
-			/*
-			 * One less used entry in the free table.
-			 */
-			be32_add_cpu(&free->hdr.nused, -1);
-			xfs_dir2_free_log_header(tp, fbp);
-			/*
-			 * If this was the last entry in the table, we can
-			 * trim the table size back.  There might be other
-			 * entries at the end referring to non-existent
-			 * data blocks, get those too.
-			 */
-			if (findex == be32_to_cpu(free->hdr.nvalid) - 1) {
-				int	i;		/* free entry index */
-
-				for (i = findex - 1;
-				     i >= 0 &&
-				     free->bests[i] == cpu_to_be16(NULLDATAOFF);
-				     i--)
-					continue;
-				free->hdr.nvalid = cpu_to_be32(i + 1);
-				logfree = 0;
-			}
-			/*
-			 * Not the last entry, just punch it out.
-			 */
-			else {
-				free->bests[findex] = cpu_to_be16(NULLDATAOFF);
-				logfree = 1;
-			}
-			/*
-			 * If there are no useful entries left in the block,
-			 * get rid of the block if we can.
-			 */
-			if (!free->hdr.nused) {
-				error = xfs_dir2_shrink_inode(args, fdb, fbp);
-				if (error == 0) {
-					fbp = NULL;
-					logfree = 0;
-				} else if (error != ENOSPC || args->total != 0)
-					return error;
-				/*
-				 * It's possible to get ENOSPC if there is no
-				 * space reservation.  In this case some one
-				 * else will eventually get rid of this block.
-				 */
-			}
-		}
-		/*
-		 * Data block is not empty, just set the free entry to
-		 * the new value.
-		 */
-		else {
-			free->bests[findex] = cpu_to_be16(longest);
-			logfree = 1;
-		}
-		/*
-		 * Log the free entry that changed, unless we got rid of it.
-		 */
-		if (logfree)
-			xfs_dir2_free_log_bests(tp, fbp, findex, findex);
+		error = xfs_dir2_data_block_free(args, hdr, free,
+						 fdb, findex, fbp, longest);
+		if (error)
+			return error;
 	}
+
 	xfs_dir2_leafn_check(dp, bp);
 	/*
 	 * Return indication of whether this leaf block is empty enough
@@ -1169,12 +1255,11 @@ xfs_dir2_leafn_toosmall(
 		/*
 		 * Read the sibling leaf block.
 		 */
-		if ((error =
-		    xfs_da_read_buf(state->args->trans, state->args->dp, blkno,
-			    -1, &bp, XFS_DATA_FORK))) {
+		error = xfs_dir2_leafn_read(state->args->trans, state->args->dp,
+					    blkno, -1, &bp);
+		if (error)
 			return error;
-		}
-		ASSERT(bp != NULL);
+
 		/*
 		 * Count bytes in the two blocks combined.
 		 */
@@ -1454,14 +1539,13 @@ xfs_dir2_node_addname_int(
 			 * This should be really rare, so there's no reason
 			 * to avoid it.
 			 */
-			if ((error = xfs_da_read_buf(tp, dp,
-					xfs_dir2_db_to_da(mp, fbno), -2, &fbp,
-					XFS_DATA_FORK))) {
+			error = xfs_dir2_free_try_read(tp, dp,
+						xfs_dir2_db_to_da(mp, fbno),
+						&fbp);
+			if (error)
 				return error;
-			}
-			if (unlikely(fbp == NULL)) {
+			if (!fbp)
 				continue;
-			}
 			free = fbp->b_addr;
 			ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
 			findex = 0;
@@ -1520,9 +1604,10 @@ xfs_dir2_node_addname_int(
 		 * that was just allocated.
 		 */
 		fbno = xfs_dir2_db_to_fdb(mp, dbno);
-		if (unlikely(error = xfs_da_read_buf(tp, dp,
-				xfs_dir2_db_to_da(mp, fbno), -2, &fbp,
-				XFS_DATA_FORK)))
+		error = xfs_dir2_free_try_read(tp, dp,
+					       xfs_dir2_db_to_da(mp, fbno),
+					       &fbp);
+		if (error)
 			return error;
 
 		/*
@@ -1561,12 +1646,12 @@ xfs_dir2_node_addname_int(
 			/*
 			 * Get a buffer for the new block.
 			 */
-			if ((error = xfs_da_get_buf(tp, dp,
-						   xfs_dir2_db_to_da(mp, fbno),
-						   -1, &fbp, XFS_DATA_FORK))) {
+			error = xfs_da_get_buf(tp, dp,
+					       xfs_dir2_db_to_da(mp, fbno),
+					       -1, &fbp, XFS_DATA_FORK);
+			if (error)
 				return error;
-			}
-			ASSERT(fbp != NULL);
+			fbp->b_ops = &xfs_dir2_free_buf_ops;
 
 			/*
 			 * Initialize the new block to be empty, and remember
@@ -1630,8 +1715,8 @@ xfs_dir2_node_addname_int(
 		/*
 		 * Read the data block in.
 		 */
-		error = xfs_da_read_buf(tp, dp, xfs_dir2_db_to_da(mp, dbno),
-				-1, &dbp, XFS_DATA_FORK);
+		error = xfs_dir2_data_read(tp, dp, xfs_dir2_db_to_da(mp, dbno),
+					   -1, &dbp);
 		if (error)
 			return error;
 		hdr = dbp->b_addr;
@@ -1917,18 +2002,15 @@ xfs_dir2_node_trim_free(
 	/*
 	 * Read the freespace block.
 	 */
-	if (unlikely(error = xfs_da_read_buf(tp, dp, (xfs_dablk_t)fo, -2, &bp,
-			XFS_DATA_FORK))) {
+	error = xfs_dir2_free_try_read(tp, dp, fo, &bp);
+	if (error)
 		return error;
-	}
-
 	/*
 	 * There can be holes in freespace.  If fo is a hole, there's
 	 * nothing to do.
 	 */
-	if (bp == NULL) {
+	if (!bp)
 		return 0;
-	}
 	free = bp->b_addr;
 	ASSERT(free->hdr.magic == cpu_to_be32(XFS_DIR2_FREE_MAGIC));
 	/*

fs/xfs/xfs_dir2_priv.h

@@ -30,6 +30,8 @@ extern int xfs_dir_cilookup_result(struct xfs_da_args *args,
 				const unsigned char *name, int len);
 
 /* xfs_dir2_block.c */
+extern const struct xfs_buf_ops xfs_dir2_block_buf_ops;
+
 extern int xfs_dir2_block_addname(struct xfs_da_args *args);
 extern int xfs_dir2_block_getdents(struct xfs_inode *dp, void *dirent,
 		xfs_off_t *offset, filldir_t filldir);
@@ -41,10 +43,19 @@ extern int xfs_dir2_leaf_to_block(struct xfs_da_args *args,
 
 /* xfs_dir2_data.c */
 #ifdef DEBUG
-extern void xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
+#define	xfs_dir2_data_check(dp,bp) __xfs_dir2_data_check(dp, bp);
 #else
 #define	xfs_dir2_data_check(dp,bp)
 #endif
+
+extern const struct xfs_buf_ops xfs_dir2_data_buf_ops;
+
+extern int __xfs_dir2_data_check(struct xfs_inode *dp, struct xfs_buf *bp);
+extern int xfs_dir2_data_read(struct xfs_trans *tp, struct xfs_inode *dp,
+		xfs_dablk_t bno, xfs_daddr_t mapped_bno, struct xfs_buf **bpp);
+extern int xfs_dir2_data_readahead(struct xfs_trans *tp, struct xfs_inode *dp,
+		xfs_dablk_t bno, xfs_daddr_t mapped_bno);
+
 extern struct xfs_dir2_data_free *
 xfs_dir2_data_freeinsert(struct xfs_dir2_data_hdr *hdr,
 		struct xfs_dir2_data_unused *dup, int *loghead);
@@ -66,6 +77,10 @@ extern void xfs_dir2_data_use_free(struct xfs_trans *tp, struct xfs_buf *bp,
 		xfs_dir2_data_aoff_t len, int *needlogp, int *needscanp);
 
 /* xfs_dir2_leaf.c */
+extern const struct xfs_buf_ops xfs_dir2_leafn_buf_ops;
+
+extern int xfs_dir2_leafn_read(struct xfs_trans *tp, struct xfs_inode *dp,
+		xfs_dablk_t fbno, xfs_daddr_t mappedbno, struct xfs_buf **bpp);
 extern int xfs_dir2_block_to_leaf(struct xfs_da_args *args,
 		struct xfs_buf *dbp);
 extern int xfs_dir2_leaf_addname(struct xfs_da_args *args);
@@ -115,6 +130,8 @@ extern int xfs_dir2_node_removename(struct xfs_da_args *args);
 extern int xfs_dir2_node_replace(struct xfs_da_args *args);
 extern int xfs_dir2_node_trim_free(struct xfs_da_args *args, xfs_fileoff_t fo,
 		int *rvalp);
+extern int xfs_dir2_free_read(struct xfs_trans *tp, struct xfs_inode *dp,
+		xfs_dablk_t fbno, struct xfs_buf **bpp);
 
 /* xfs_dir2_sf.c */
 extern xfs_ino_t xfs_dir2_sf_get_parent_ino(struct xfs_dir2_sf_hdr *sfp);

fs/xfs/xfs_dquot.c

@@ -248,7 +248,59 @@ xfs_qm_init_dquot_blk(
 	xfs_trans_log_buf(tp, bp, 0, BBTOB(q->qi_dqchunklen) - 1);
 }
 
+static void
+xfs_dquot_buf_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_dqblk	*d = (struct xfs_dqblk *)bp->b_addr;
+	struct xfs_disk_dquot	*ddq;
+	xfs_dqid_t		id = 0;
+	int			i;
+
+	/*
+	 * On the first read of the buffer, verify that each dquot is valid.
+	 * We don't know what the id of the dquot is supposed to be, just that
+	 * they should be increasing monotonically within the buffer. If the
+	 * first id is corrupt, then it will fail on the second dquot in the
+	 * buffer so corruptions could point to the wrong dquot in this case.
+	 */
+	for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
+		int	error;
+
+		ddq = &d[i].dd_diskdq;
+
+		if (i == 0)
+			id = be32_to_cpu(ddq->d_id);
+
+		error = xfs_qm_dqcheck(mp, ddq, id + i, 0, XFS_QMOPT_DOWARN,
+					"xfs_dquot_read_verify");
+		if (error) {
+			XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, d);
+			xfs_buf_ioerror(bp, EFSCORRUPTED);
+			break;
+		}
+	}
+}
+
+static void
+xfs_dquot_buf_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dquot_buf_verify(bp);
+}
+
+void
+xfs_dquot_buf_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_dquot_buf_verify(bp);
+}
 
+const struct xfs_buf_ops xfs_dquot_buf_ops = {
+	.verify_read = xfs_dquot_buf_read_verify,
+	.verify_write = xfs_dquot_buf_write_verify,
+};
 
 /*
  * Allocate a block and fill it with dquots.
@@ -315,6 +367,7 @@ xfs_qm_dqalloc(
 	error = xfs_buf_geterror(bp);
 	if (error)
 		goto error1;
+	bp->b_ops = &xfs_dquot_buf_ops;
 
 	/*
 	 * Make a chunk of dquots out of this buffer and log
@@ -359,6 +412,51 @@ xfs_qm_dqalloc(
 
 	return (error);
 }
+STATIC int
+xfs_qm_dqrepair(
+	struct xfs_mount	*mp,
+	struct xfs_trans	*tp,
+	struct xfs_dquot	*dqp,
+	xfs_dqid_t		firstid,
+	struct xfs_buf		**bpp)
+{
+	int			error;
+	struct xfs_disk_dquot	*ddq;
+	struct xfs_dqblk	*d;
+	int			i;
+
+	/*
+	 * Read the buffer without verification so we get the corrupted
+	 * buffer returned to us. make sure we verify it on write, though.
+	 */
+	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, dqp->q_blkno,
+				   mp->m_quotainfo->qi_dqchunklen,
+				   0, bpp, NULL);
+
+	if (error) {
+		ASSERT(*bpp == NULL);
+		return XFS_ERROR(error);
+	}
+	(*bpp)->b_ops = &xfs_dquot_buf_ops;
+
+	ASSERT(xfs_buf_islocked(*bpp));
+	d = (struct xfs_dqblk *)(*bpp)->b_addr;
+
+	/* Do the actual repair of dquots in this buffer */
+	for (i = 0; i < mp->m_quotainfo->qi_dqperchunk; i++) {
+		ddq = &d[i].dd_diskdq;
+		error = xfs_qm_dqcheck(mp, ddq, firstid + i,
+				       dqp->dq_flags & XFS_DQ_ALLTYPES,
+				       XFS_QMOPT_DQREPAIR, "xfs_qm_dqrepair");
+		if (error) {
+			/* repair failed, we're screwed */
+			xfs_trans_brelse(tp, *bpp);
+			return XFS_ERROR(EIO);
+		}
+	}
+
+	return 0;
+}
 
 /*
  * Maps a dquot to the buffer containing its on-disk version.
@@ -378,7 +476,6 @@ xfs_qm_dqtobp(
 	xfs_buf_t	*bp;
 	xfs_inode_t	*quotip = XFS_DQ_TO_QIP(dqp);
 	xfs_mount_t	*mp = dqp->q_mount;
-	xfs_disk_dquot_t *ddq;
 	xfs_dqid_t	id = be32_to_cpu(dqp->q_core.d_id);
 	xfs_trans_t	*tp = (tpp ? *tpp : NULL);
 
@@ -439,33 +536,24 @@ xfs_qm_dqtobp(
 		error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
 					   dqp->q_blkno,
 					   mp->m_quotainfo->qi_dqchunklen,
-					   0, &bp);
-		if (error || !bp)
-			return XFS_ERROR(error);
-	}
-
-	ASSERT(xfs_buf_islocked(bp));
+					   0, &bp, &xfs_dquot_buf_ops);
 
-	/*
-	 * calculate the location of the dquot inside the buffer.
-	 */
-	ddq = bp->b_addr + dqp->q_bufoffset;
+		if (error == EFSCORRUPTED && (flags & XFS_QMOPT_DQREPAIR)) {
+			xfs_dqid_t firstid = (xfs_dqid_t)map.br_startoff *
+						mp->m_quotainfo->qi_dqperchunk;
+			ASSERT(bp == NULL);
+			error = xfs_qm_dqrepair(mp, tp, dqp, firstid, &bp);
+		}
 
-	/*
-	 * A simple sanity check in case we got a corrupted dquot...
-	 */
-	error = xfs_qm_dqcheck(mp, ddq, id, dqp->dq_flags & XFS_DQ_ALLTYPES,
-			   flags & (XFS_QMOPT_DQREPAIR|XFS_QMOPT_DOWARN),
-			   "dqtobp");
-	if (error) {
-		if (!(flags & XFS_QMOPT_DQREPAIR)) {
-			xfs_trans_brelse(tp, bp);
-			return XFS_ERROR(EIO);
+		if (error) {
+			ASSERT(bp == NULL);
+			return XFS_ERROR(error);
 		}
 	}
 
+	ASSERT(xfs_buf_islocked(bp));
 	*O_bpp = bp;
-	*O_ddpp = ddq;
+	*O_ddpp = bp->b_addr + dqp->q_bufoffset;
 
 	return (0);
 }
@@ -920,7 +1008,7 @@ xfs_qm_dqflush(
 	 * Get the buffer containing the on-disk dquot
 	 */
 	error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dqp->q_blkno,
-				   mp->m_quotainfo->qi_dqchunklen, 0, &bp);
+				   mp->m_quotainfo->qi_dqchunklen, 0, &bp, NULL);
 	if (error)
 		goto out_unlock;
 

fs/xfs/xfs_dquot.h

@@ -161,4 +161,6 @@ static inline struct xfs_dquot *xfs_qm_dqhold(struct xfs_dquot *dqp)
 	return dqp;
 }
 
+extern const struct xfs_buf_ops xfs_dquot_buf_ops;
+
 #endif /* __XFS_DQUOT_H__ */

fs/xfs/xfs_export.c

@@ -29,6 +29,7 @@
 #include "xfs_inode.h"
 #include "xfs_inode_item.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 /*
  * Note that we only accept fileids which are long enough rather than allow

fs/xfs/xfs_file.c

@@ -31,6 +31,8 @@
 #include "xfs_error.h"
 #include "xfs_vnodeops.h"
 #include "xfs_da_btree.h"
+#include "xfs_dir2_format.h"
+#include "xfs_dir2_priv.h"
 #include "xfs_ioctl.h"
 #include "xfs_trace.h"
 
@@ -84,7 +86,7 @@ xfs_rw_ilock_demote(
  *	valid before the operation, it will be read from disk before
  *	being partially zeroed.
  */
-STATIC int
+int
 xfs_iozero(
 	struct xfs_inode	*ip,	/* inode			*/
 	loff_t			pos,	/* offset in file		*/
@@ -255,15 +257,14 @@ xfs_file_aio_read(
 		xfs_buftarg_t	*target =
 			XFS_IS_REALTIME_INODE(ip) ?
 				mp->m_rtdev_targp : mp->m_ddev_targp;
-		if ((iocb->ki_pos & target->bt_smask) ||
-		    (size & target->bt_smask)) {
-			if (iocb->ki_pos == i_size_read(inode))
+		if ((pos & target->bt_smask) || (size & target->bt_smask)) {
+			if (pos == i_size_read(inode))
 				return 0;
 			return -XFS_ERROR(EINVAL);
 		}
 	}
 
-	n = mp->m_super->s_maxbytes - iocb->ki_pos;
+	n = mp->m_super->s_maxbytes - pos;
 	if (n <= 0 || size == 0)
 		return 0;
 
@@ -289,20 +290,21 @@ xfs_file_aio_read(
 		xfs_rw_ilock(ip, XFS_IOLOCK_EXCL);
 
 		if (inode->i_mapping->nrpages) {
-			ret = -xfs_flushinval_pages(ip,
-					(iocb->ki_pos & PAGE_CACHE_MASK),
-					-1, FI_REMAPF_LOCKED);
+			ret = -filemap_write_and_wait_range(
+							VFS_I(ip)->i_mapping,
+							pos, -1);
 			if (ret) {
 				xfs_rw_iunlock(ip, XFS_IOLOCK_EXCL);
 				return ret;
 			}
+			truncate_pagecache_range(VFS_I(ip), pos, -1);
 		}
 		xfs_rw_ilock_demote(ip, XFS_IOLOCK_EXCL);
 	}
 
-	trace_xfs_file_read(ip, size, iocb->ki_pos, ioflags);
+	trace_xfs_file_read(ip, size, pos, ioflags);
 
-	ret = generic_file_aio_read(iocb, iovp, nr_segs, iocb->ki_pos);
+	ret = generic_file_aio_read(iocb, iovp, nr_segs, pos);
 	if (ret > 0)
 		XFS_STATS_ADD(xs_read_bytes, ret);
 
@@ -670,10 +672,11 @@ xfs_file_dio_aio_write(
 		goto out;
 
 	if (mapping->nrpages) {
-		ret = -xfs_flushinval_pages(ip, (pos & PAGE_CACHE_MASK), -1,
-							FI_REMAPF_LOCKED);
+		ret = -filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
+						    pos, -1);
 		if (ret)
 			goto out;
+		truncate_pagecache_range(VFS_I(ip), pos, -1);
 	}
 
 	/*
@@ -728,16 +731,17 @@ xfs_file_buffered_aio_write(
 write_retry:
 	trace_xfs_file_buffered_write(ip, count, iocb->ki_pos, 0);
 	ret = generic_file_buffered_write(iocb, iovp, nr_segs,
-			pos, &iocb->ki_pos, count, ret);
+			pos, &iocb->ki_pos, count, 0);
+
 	/*
-	 * if we just got an ENOSPC, flush the inode now we aren't holding any
-	 * page locks and retry *once*
+	 * If we just got an ENOSPC, try to write back all dirty inodes to
+	 * convert delalloc space to free up some of the excess reserved
+	 * metadata space.
 	 */
 	if (ret == -ENOSPC && !enospc) {
 		enospc = 1;
-		ret = -xfs_flush_pages(ip, 0, -1, 0, FI_NONE);
-		if (!ret)
-			goto write_retry;
+		xfs_flush_inodes(ip->i_mount);
+		goto write_retry;
 	}
 
 	current->backing_dev_info = NULL;
@@ -889,7 +893,7 @@ xfs_dir_open(
 	 */
 	mode = xfs_ilock_map_shared(ip);
 	if (ip->i_d.di_nextents > 0)
-		xfs_da_reada_buf(NULL, ip, 0, XFS_DATA_FORK);
+		xfs_dir2_data_readahead(NULL, ip, 0, -1);
 	xfs_iunlock(ip, mode);
 	return 0;
 }

fs/xfs/xfs_fs.h

@@ -233,7 +233,8 @@ typedef struct xfs_fsop_resblks {
 #define XFS_FSOP_GEOM_FLAGS_LOGV2	0x0100	/* log format version 2	*/
 #define XFS_FSOP_GEOM_FLAGS_SECTOR	0x0200	/* sector sizes >1BB	*/
 #define XFS_FSOP_GEOM_FLAGS_ATTR2	0x0400	/* inline attributes rework */
-#define XFS_FSOP_GEOM_FLAGS_DIRV2CI	0x1000	/* ASCII only CI names */
+#define XFS_FSOP_GEOM_FLAGS_PROJID32	0x0800  /* 32-bit project IDs	*/
+#define XFS_FSOP_GEOM_FLAGS_DIRV2CI	0x1000	/* ASCII only CI names	*/
 #define XFS_FSOP_GEOM_FLAGS_LAZYSB	0x4000	/* lazy superblock counters */
 
 
@@ -338,6 +339,35 @@ typedef struct xfs_error_injection {
 } xfs_error_injection_t;
 
 
+/*
+ * Speculative preallocation trimming.
+ */
+#define XFS_EOFBLOCKS_VERSION		1
+struct xfs_eofblocks {
+	__u32		eof_version;
+	__u32		eof_flags;
+	uid_t		eof_uid;
+	gid_t		eof_gid;
+	prid_t		eof_prid;
+	__u32		pad32;
+	__u64		eof_min_file_size;
+	__u64		pad64[12];
+};
+
+/* eof_flags values */
+#define XFS_EOF_FLAGS_SYNC		(1 << 0) /* sync/wait mode scan */
+#define XFS_EOF_FLAGS_UID		(1 << 1) /* filter by uid */
+#define XFS_EOF_FLAGS_GID		(1 << 2) /* filter by gid */
+#define XFS_EOF_FLAGS_PRID		(1 << 3) /* filter by project id */
+#define XFS_EOF_FLAGS_MINFILESIZE	(1 << 4) /* filter by min file size */
+#define XFS_EOF_FLAGS_VALID	\
+	(XFS_EOF_FLAGS_SYNC |	\
+	 XFS_EOF_FLAGS_UID |	\
+	 XFS_EOF_FLAGS_GID |	\
+	 XFS_EOF_FLAGS_PRID |	\
+	 XFS_EOF_FLAGS_MINFILESIZE)
+
+
 /*
  * The user-level Handle Request interface structure.
  */
@@ -456,6 +486,7 @@ typedef struct xfs_handle {
 /*	XFS_IOC_GETBIOSIZE ---- deprecated 47	   */
 #define XFS_IOC_GETBMAPX	_IOWR('X', 56, struct getbmap)
 #define XFS_IOC_ZERO_RANGE	_IOW ('X', 57, struct xfs_flock64)
+#define XFS_IOC_FREE_EOFBLOCKS	_IOR ('X', 58, struct xfs_eofblocks)
 
 /*
  * ioctl commands that replace IRIX syssgi()'s

fs/xfs/xfs_fs_subr.c

@@ -1,96 +0,0 @@
-/*
- * Copyright (c) 2000-2002,2005-2006 Silicon Graphics, Inc.
- * All Rights Reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it would be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write the Free Software Foundation,
- * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
- */
-#include "xfs.h"
-#include "xfs_vnodeops.h"
-#include "xfs_bmap_btree.h"
-#include "xfs_inode.h"
-#include "xfs_trace.h"
-
-/*
- * note: all filemap functions return negative error codes. These
- * need to be inverted before returning to the xfs core functions.
- */
-void
-xfs_tosspages(
-	xfs_inode_t	*ip,
-	xfs_off_t	first,
-	xfs_off_t	last,
-	int		fiopt)
-{
-	/* can't toss partial tail pages, so mask them out */
-	last &= ~(PAGE_SIZE - 1);
-	truncate_inode_pages_range(VFS_I(ip)->i_mapping, first, last - 1);
-}
-
-int
-xfs_flushinval_pages(
-	xfs_inode_t	*ip,
-	xfs_off_t	first,
-	xfs_off_t	last,
-	int		fiopt)
-{
-	struct address_space *mapping = VFS_I(ip)->i_mapping;
-	int		ret = 0;
-
-	trace_xfs_pagecache_inval(ip, first, last);
-
-	xfs_iflags_clear(ip, XFS_ITRUNCATED);
-	ret = filemap_write_and_wait_range(mapping, first,
-				last == -1 ? LLONG_MAX : last);
-	if (!ret)
-		truncate_inode_pages_range(mapping, first, last);
-	return -ret;
-}
-
-int
-xfs_flush_pages(
-	xfs_inode_t	*ip,
-	xfs_off_t	first,
-	xfs_off_t	last,
-	uint64_t	flags,
-	int		fiopt)
-{
-	struct address_space *mapping = VFS_I(ip)->i_mapping;
-	int		ret = 0;
-	int		ret2;
-
-	xfs_iflags_clear(ip, XFS_ITRUNCATED);
-	ret = -filemap_fdatawrite_range(mapping, first,
-				last == -1 ? LLONG_MAX : last);
-	if (flags & XBF_ASYNC)
-		return ret;
-	ret2 = xfs_wait_on_pages(ip, first, last);
-	if (!ret)
-		ret = ret2;
-	return ret;
-}
-
-int
-xfs_wait_on_pages(
-	xfs_inode_t	*ip,
-	xfs_off_t	first,
-	xfs_off_t	last)
-{
-	struct address_space *mapping = VFS_I(ip)->i_mapping;
-
-	if (mapping_tagged(mapping, PAGECACHE_TAG_WRITEBACK)) {
-		return -filemap_fdatawait_range(mapping, first,
-					last == -1 ? XFS_ISIZE(ip) - 1 : last);
-	}
-	return 0;
-}

fs/xfs/xfs_fsops.c

@@ -97,7 +97,9 @@ xfs_fs_geometry(
 			(xfs_sb_version_haslazysbcount(&mp->m_sb) ?
 				XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
 			(xfs_sb_version_hasattr2(&mp->m_sb) ?
-				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0);
+				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
+			(xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
+				XFS_FSOP_GEOM_FLAGS_PROJID32 : 0);
 		geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
 				mp->m_sb.sb_logsectsize : BBSIZE;
 		geo->rtsectsize = mp->m_sb.sb_blocksize;
@@ -112,18 +114,40 @@ xfs_fs_geometry(
 	return 0;
 }
 
+static struct xfs_buf *
+xfs_growfs_get_hdr_buf(
+	struct xfs_mount	*mp,
+	xfs_daddr_t		blkno,
+	size_t			numblks,
+	int			flags,
+	const struct xfs_buf_ops *ops)
+{
+	struct xfs_buf		*bp;
+
+	bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
+	if (!bp)
+		return NULL;
+
+	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
+	bp->b_bn = blkno;
+	bp->b_maps[0].bm_bn = blkno;
+	bp->b_ops = ops;
+
+	return bp;
+}
+
 static int
 xfs_growfs_data_private(
 	xfs_mount_t		*mp,		/* mount point for filesystem */
 	xfs_growfs_data_t	*in)		/* growfs data input struct */
 {
 	xfs_agf_t		*agf;
+	struct xfs_agfl		*agfl;
 	xfs_agi_t		*agi;
 	xfs_agnumber_t		agno;
 	xfs_extlen_t		agsize;
 	xfs_extlen_t		tmpsize;
 	xfs_alloc_rec_t		*arec;
-	struct xfs_btree_block	*block;
 	xfs_buf_t		*bp;
 	int			bucket;
 	int			dpct;
@@ -146,9 +170,14 @@ xfs_growfs_data_private(
 	dpct = pct - mp->m_sb.sb_imax_pct;
 	bp = xfs_buf_read_uncached(mp->m_ddev_targp,
 				XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
-				XFS_FSS_TO_BB(mp, 1), 0);
+				XFS_FSS_TO_BB(mp, 1), 0, NULL);
 	if (!bp)
 		return EIO;
+	if (bp->b_error) {
+		int	error = bp->b_error;
+		xfs_buf_relse(bp);
+		return error;
+	}
 	xfs_buf_relse(bp);
 
 	new = nb;	/* use new as a temporary here */
@@ -186,17 +215,18 @@ xfs_growfs_data_private(
 	nfree = 0;
 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
 		/*
-		 * AG freelist header block
+		 * AG freespace header block
 		 */
-		bp = xfs_buf_get(mp->m_ddev_targp,
-				 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
-				 XFS_FSS_TO_BB(mp, 1), 0);
+		bp = xfs_growfs_get_hdr_buf(mp,
+				XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
+				XFS_FSS_TO_BB(mp, 1), 0,
+				&xfs_agf_buf_ops);
 		if (!bp) {
 			error = ENOMEM;
 			goto error0;
 		}
+
 		agf = XFS_BUF_TO_AGF(bp);
-		memset(agf, 0, mp->m_sb.sb_sectsize);
 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
 		agf->agf_seqno = cpu_to_be32(agno);
@@ -222,18 +252,40 @@ xfs_growfs_data_private(
 		if (error)
 			goto error0;
 
+		/*
+		 * AG freelist header block
+		 */
+		bp = xfs_growfs_get_hdr_buf(mp,
+				XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
+				XFS_FSS_TO_BB(mp, 1), 0,
+				&xfs_agfl_buf_ops);
+		if (!bp) {
+			error = ENOMEM;
+			goto error0;
+		}
+
+		agfl = XFS_BUF_TO_AGFL(bp);
+		for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
+			agfl->agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
+
+		error = xfs_bwrite(bp);
+		xfs_buf_relse(bp);
+		if (error)
+			goto error0;
+
 		/*
 		 * AG inode header block
 		 */
-		bp = xfs_buf_get(mp->m_ddev_targp,
-				 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
-				 XFS_FSS_TO_BB(mp, 1), 0);
+		bp = xfs_growfs_get_hdr_buf(mp,
+				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
+				XFS_FSS_TO_BB(mp, 1), 0,
+				&xfs_agi_buf_ops);
 		if (!bp) {
 			error = ENOMEM;
 			goto error0;
 		}
+
 		agi = XFS_BUF_TO_AGI(bp);
-		memset(agi, 0, mp->m_sb.sb_sectsize);
 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
 		agi->agi_seqno = cpu_to_be32(agno);
@@ -254,24 +306,22 @@ xfs_growfs_data_private(
 		/*
 		 * BNO btree root block
 		 */
-		bp = xfs_buf_get(mp->m_ddev_targp,
-				 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
-				 BTOBB(mp->m_sb.sb_blocksize), 0);
+		bp = xfs_growfs_get_hdr_buf(mp,
+				XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
+				BTOBB(mp->m_sb.sb_blocksize), 0,
+				&xfs_allocbt_buf_ops);
+
 		if (!bp) {
 			error = ENOMEM;
 			goto error0;
 		}
-		block = XFS_BUF_TO_BLOCK(bp);
-		memset(block, 0, mp->m_sb.sb_blocksize);
-		block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
-		block->bb_level = 0;
-		block->bb_numrecs = cpu_to_be16(1);
-		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
-		arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
+
+		xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1, 0);
+		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
 		arec->ar_blockcount = cpu_to_be32(
 			agsize - be32_to_cpu(arec->ar_startblock));
+
 		error = xfs_bwrite(bp);
 		xfs_buf_relse(bp);
 		if (error)
@@ -280,25 +330,22 @@ xfs_growfs_data_private(
 		/*
 		 * CNT btree root block
 		 */
-		bp = xfs_buf_get(mp->m_ddev_targp,
-				 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
-				 BTOBB(mp->m_sb.sb_blocksize), 0);
+		bp = xfs_growfs_get_hdr_buf(mp,
+				XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
+				BTOBB(mp->m_sb.sb_blocksize), 0,
+				&xfs_allocbt_buf_ops);
 		if (!bp) {
 			error = ENOMEM;
 			goto error0;
 		}
-		block = XFS_BUF_TO_BLOCK(bp);
-		memset(block, 0, mp->m_sb.sb_blocksize);
-		block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
-		block->bb_level = 0;
-		block->bb_numrecs = cpu_to_be16(1);
-		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
-		arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
+
+		xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1, 0);
+		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
 		arec->ar_blockcount = cpu_to_be32(
 			agsize - be32_to_cpu(arec->ar_startblock));
 		nfree += be32_to_cpu(arec->ar_blockcount);
+
 		error = xfs_bwrite(bp);
 		xfs_buf_relse(bp);
 		if (error)
@@ -307,20 +354,17 @@ xfs_growfs_data_private(
 		/*
 		 * INO btree root block
 		 */
-		bp = xfs_buf_get(mp->m_ddev_targp,
-				 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
-				 BTOBB(mp->m_sb.sb_blocksize), 0);
+		bp = xfs_growfs_get_hdr_buf(mp,
+				XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
+				BTOBB(mp->m_sb.sb_blocksize), 0,
+				&xfs_inobt_buf_ops);
 		if (!bp) {
 			error = ENOMEM;
 			goto error0;
 		}
-		block = XFS_BUF_TO_BLOCK(bp);
-		memset(block, 0, mp->m_sb.sb_blocksize);
-		block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
-		block->bb_level = 0;
-		block->bb_numrecs = 0;
-		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
-		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
+
+		xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0, 0);
+
 		error = xfs_bwrite(bp);
 		xfs_buf_relse(bp);
 		if (error)
@@ -408,14 +452,16 @@ xfs_growfs_data_private(
 		if (agno < oagcount) {
 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
-				  XFS_FSS_TO_BB(mp, 1), 0, &bp);
+				  XFS_FSS_TO_BB(mp, 1), 0, &bp,
+				  &xfs_sb_buf_ops);
 		} else {
 			bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
 				  XFS_FSS_TO_BB(mp, 1), 0);
-			if (bp)
+			if (bp) {
+				bp->b_ops = &xfs_sb_buf_ops;
 				xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
-			else
+			} else
 				error = ENOMEM;
 		}
 
@@ -426,6 +472,7 @@ xfs_growfs_data_private(
 			break;
 		}
 		xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
+
 		/*
 		 * If we get an error writing out the alternate superblocks,
 		 * just issue a warning and continue.  The real work is

fs/xfs/xfs_globals.c

@@ -21,7 +21,8 @@
 /*
  * Tunable XFS parameters.  xfs_params is required even when CONFIG_SYSCTL=n,
  * other XFS code uses these values.  Times are measured in centisecs (i.e.
- * 100ths of a second).
+ * 100ths of a second) with the exception of eofb_timer, which is measured in
+ * seconds.
  */
 xfs_param_t xfs_params = {
 			  /*	MIN		DFLT		MAX	*/
@@ -40,4 +41,5 @@ xfs_param_t xfs_params = {
 	.rotorstep	= {	1,		1,		255	},
 	.inherit_nodfrg	= {	0,		1,		1	},
 	.fstrm_timer	= {	1,		30*100,		3600*100},
+	.eofb_timer	= {	1,		300,		3600*24},
 };

fs/xfs/xfs_ialloc.c

@@ -200,7 +200,8 @@ xfs_ialloc_inode_init(
 		 */
 		d = XFS_AGB_TO_DADDR(mp, agno, agbno + (j * blks_per_cluster));
 		fbuf = xfs_trans_get_buf(tp, mp->m_ddev_targp, d,
-					 mp->m_bsize * blks_per_cluster, 0);
+					 mp->m_bsize * blks_per_cluster,
+					 XBF_UNMAPPED);
 		if (!fbuf)
 			return ENOMEM;
 		/*
@@ -210,6 +211,7 @@ xfs_ialloc_inode_init(
 		 *	to log a whole cluster of inodes instead of all the
 		 *	individual transactions causing a lot of log traffic.
 		 */
+		fbuf->b_ops = &xfs_inode_buf_ops;
 		xfs_buf_zero(fbuf, 0, ninodes << mp->m_sb.sb_inodelog);
 		for (i = 0; i < ninodes; i++) {
 			int	ioffset = i << mp->m_sb.sb_inodelog;
@@ -877,9 +879,9 @@ error0:
  * This function is designed to be called twice if it has to do an allocation
  * to make more free inodes.  On the first call, *IO_agbp should be set to NULL.
  * If an inode is available without having to performn an allocation, an inode
- * number is returned.  In this case, *IO_agbp would be NULL.  If an allocation
- * needes to be done, xfs_dialloc would return the current AGI buffer in
- * *IO_agbp.  The caller should then commit the current transaction, allocate a
+ * number is returned.  In this case, *IO_agbp is set to NULL.  If an allocation
+ * needs to be done, xfs_dialloc returns the current AGI buffer in *IO_agbp.
+ * The caller should then commit the current transaction, allocate a
  * new transaction, and call xfs_dialloc() again, passing in the previous value
  * of *IO_agbp.  IO_agbp should be held across the transactions. Since the AGI
  * buffer is locked across the two calls, the second call is guaranteed to have
@@ -1472,6 +1474,57 @@ xfs_check_agi_unlinked(
 #define xfs_check_agi_unlinked(agi)
 #endif
 
+static void
+xfs_agi_verify(
+	struct xfs_buf	*bp)
+{
+	struct xfs_mount *mp = bp->b_target->bt_mount;
+	struct xfs_agi	*agi = XFS_BUF_TO_AGI(bp);
+	int		agi_ok;
+
+	/*
+	 * Validate the magic number of the agi block.
+	 */
+	agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
+		XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum));
+
+	/*
+	 * during growfs operations, the perag is not fully initialised,
+	 * so we can't use it for any useful checking. growfs ensures we can't
+	 * use it by using uncached buffers that don't have the perag attached
+	 * so we can detect and avoid this problem.
+	 */
+	if (bp->b_pag)
+		agi_ok = agi_ok && be32_to_cpu(agi->agi_seqno) ==
+						bp->b_pag->pag_agno;
+
+	if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
+			XFS_RANDOM_IALLOC_READ_AGI))) {
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, agi);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+	xfs_check_agi_unlinked(agi);
+}
+
+static void
+xfs_agi_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_agi_verify(bp);
+}
+
+static void
+xfs_agi_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_agi_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_agi_buf_ops = {
+	.verify_read = xfs_agi_read_verify,
+	.verify_write = xfs_agi_write_verify,
+};
+
 /*
  * Read in the allocation group header (inode allocation section)
  */
@@ -1482,38 +1535,18 @@ xfs_read_agi(
 	xfs_agnumber_t		agno,	/* allocation group number */
 	struct xfs_buf		**bpp)	/* allocation group hdr buf */
 {
-	struct xfs_agi		*agi;	/* allocation group header */
-	int			agi_ok;	/* agi is consistent */
 	int			error;
 
 	ASSERT(agno != NULLAGNUMBER);
 
 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
 			XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
-			XFS_FSS_TO_BB(mp, 1), 0, bpp);
+			XFS_FSS_TO_BB(mp, 1), 0, bpp, &xfs_agi_buf_ops);
 	if (error)
 		return error;
 
 	ASSERT(!xfs_buf_geterror(*bpp));
-	agi = XFS_BUF_TO_AGI(*bpp);
-
-	/*
-	 * Validate the magic number of the agi block.
-	 */
-	agi_ok = agi->agi_magicnum == cpu_to_be32(XFS_AGI_MAGIC) &&
-		XFS_AGI_GOOD_VERSION(be32_to_cpu(agi->agi_versionnum)) &&
-		be32_to_cpu(agi->agi_seqno) == agno;
-	if (unlikely(XFS_TEST_ERROR(!agi_ok, mp, XFS_ERRTAG_IALLOC_READ_AGI,
-			XFS_RANDOM_IALLOC_READ_AGI))) {
-		XFS_CORRUPTION_ERROR("xfs_read_agi", XFS_ERRLEVEL_LOW,
-				     mp, agi);
-		xfs_trans_brelse(tp, *bpp);
-		return XFS_ERROR(EFSCORRUPTED);
-	}
-
 	xfs_buf_set_ref(*bpp, XFS_AGI_REF);
-
-	xfs_check_agi_unlinked(agi);
 	return 0;
 }
 

fs/xfs/xfs_ialloc.h

@@ -147,7 +147,9 @@ int xfs_inobt_lookup(struct xfs_btree_cur *cur, xfs_agino_t ino,
 /*
  * Get the data from the pointed-to record.
  */
-extern int xfs_inobt_get_rec(struct xfs_btree_cur *cur,
+int xfs_inobt_get_rec(struct xfs_btree_cur *cur,
 		xfs_inobt_rec_incore_t *rec, int *stat);
 
+extern const struct xfs_buf_ops xfs_agi_buf_ops;
+
 #endif	/* __XFS_IALLOC_H__ */

fs/xfs/xfs_ialloc_btree.c

@@ -33,6 +33,7 @@
 #include "xfs_ialloc.h"
 #include "xfs_alloc.h"
 #include "xfs_error.h"
+#include "xfs_trace.h"
 
 
 STATIC int
@@ -181,6 +182,59 @@ xfs_inobt_key_diff(
 			  cur->bc_rec.i.ir_startino;
 }
 
+void
+xfs_inobt_verify(
+	struct xfs_buf		*bp)
+{
+	struct xfs_mount	*mp = bp->b_target->bt_mount;
+	struct xfs_btree_block	*block = XFS_BUF_TO_BLOCK(bp);
+	unsigned int		level;
+	int			sblock_ok; /* block passes checks */
+
+	/* magic number and level verification */
+	level = be16_to_cpu(block->bb_level);
+	sblock_ok = block->bb_magic == cpu_to_be32(XFS_IBT_MAGIC) &&
+		    level < mp->m_in_maxlevels;
+
+	/* numrecs verification */
+	sblock_ok = sblock_ok &&
+		be16_to_cpu(block->bb_numrecs) <= mp->m_inobt_mxr[level != 0];
+
+	/* sibling pointer verification */
+	sblock_ok = sblock_ok &&
+		(block->bb_u.s.bb_leftsib == cpu_to_be32(NULLAGBLOCK) ||
+		 be32_to_cpu(block->bb_u.s.bb_leftsib) < mp->m_sb.sb_agblocks) &&
+		block->bb_u.s.bb_leftsib &&
+		(block->bb_u.s.bb_rightsib == cpu_to_be32(NULLAGBLOCK) ||
+		 be32_to_cpu(block->bb_u.s.bb_rightsib) < mp->m_sb.sb_agblocks) &&
+		block->bb_u.s.bb_rightsib;
+
+	if (!sblock_ok) {
+		trace_xfs_btree_corrupt(bp, _RET_IP_);
+		XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp, block);
+		xfs_buf_ioerror(bp, EFSCORRUPTED);
+	}
+}
+
+static void
+xfs_inobt_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_inobt_verify(bp);
+}
+
+static void
+xfs_inobt_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_inobt_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_inobt_buf_ops = {
+	.verify_read = xfs_inobt_read_verify,
+	.verify_write = xfs_inobt_write_verify,
+};
+
 #ifdef DEBUG
 STATIC int
 xfs_inobt_keys_inorder(
@@ -218,6 +272,7 @@ static const struct xfs_btree_ops xfs_inobt_ops = {
 	.init_rec_from_cur	= xfs_inobt_init_rec_from_cur,
 	.init_ptr_from_cur	= xfs_inobt_init_ptr_from_cur,
 	.key_diff		= xfs_inobt_key_diff,
+	.buf_ops		= &xfs_inobt_buf_ops,
 #ifdef DEBUG
 	.keys_inorder		= xfs_inobt_keys_inorder,
 	.recs_inorder		= xfs_inobt_recs_inorder,

fs/xfs/xfs_ialloc_btree.h

@@ -109,4 +109,6 @@ extern struct xfs_btree_cur *xfs_inobt_init_cursor(struct xfs_mount *,
 		struct xfs_trans *, struct xfs_buf *, xfs_agnumber_t);
 extern int xfs_inobt_maxrecs(struct xfs_mount *, int, int);
 
+extern const struct xfs_buf_ops xfs_inobt_buf_ops;
+
 #endif	/* __XFS_IALLOC_BTREE_H__ */

fs/xfs/xfs_sync.c → fs/xfs/xfs_icache.c

@@ -19,6 +19,7 @@
 #include "xfs_fs.h"
 #include "xfs_types.h"
 #include "xfs_log.h"
+#include "xfs_log_priv.h"
 #include "xfs_inum.h"
 #include "xfs_trans.h"
 #include "xfs_trans_priv.h"
@@ -35,11 +36,425 @@
 #include "xfs_quota.h"
 #include "xfs_trace.h"
 #include "xfs_fsops.h"
+#include "xfs_icache.h"
 
 #include <linux/kthread.h>
 #include <linux/freezer.h>
 
-struct workqueue_struct	*xfs_syncd_wq;	/* sync workqueue */
+STATIC void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp,
+				struct xfs_perag *pag, struct xfs_inode *ip);
+
+/*
+ * Allocate and initialise an xfs_inode.
+ */
+STATIC struct xfs_inode *
+xfs_inode_alloc(
+	struct xfs_mount	*mp,
+	xfs_ino_t		ino)
+{
+	struct xfs_inode	*ip;
+
+	/*
+	 * if this didn't occur in transactions, we could use
+	 * KM_MAYFAIL and return NULL here on ENOMEM. Set the
+	 * code up to do this anyway.
+	 */
+	ip = kmem_zone_alloc(xfs_inode_zone, KM_SLEEP);
+	if (!ip)
+		return NULL;
+	if (inode_init_always(mp->m_super, VFS_I(ip))) {
+		kmem_zone_free(xfs_inode_zone, ip);
+		return NULL;
+	}
+
+	ASSERT(atomic_read(&ip->i_pincount) == 0);
+	ASSERT(!spin_is_locked(&ip->i_flags_lock));
+	ASSERT(!xfs_isiflocked(ip));
+	ASSERT(ip->i_ino == 0);
+
+	mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
+
+	/* initialise the xfs inode */
+	ip->i_ino = ino;
+	ip->i_mount = mp;
+	memset(&ip->i_imap, 0, sizeof(struct xfs_imap));
+	ip->i_afp = NULL;
+	memset(&ip->i_df, 0, sizeof(xfs_ifork_t));
+	ip->i_flags = 0;
+	ip->i_delayed_blks = 0;
+	memset(&ip->i_d, 0, sizeof(xfs_icdinode_t));
+
+	return ip;
+}
+
+STATIC void
+xfs_inode_free_callback(
+	struct rcu_head		*head)
+{
+	struct inode		*inode = container_of(head, struct inode, i_rcu);
+	struct xfs_inode	*ip = XFS_I(inode);
+
+	kmem_zone_free(xfs_inode_zone, ip);
+}
+
+STATIC void
+xfs_inode_free(
+	struct xfs_inode	*ip)
+{
+	switch (ip->i_d.di_mode & S_IFMT) {
+	case S_IFREG:
+	case S_IFDIR:
+	case S_IFLNK:
+		xfs_idestroy_fork(ip, XFS_DATA_FORK);
+		break;
+	}
+
+	if (ip->i_afp)
+		xfs_idestroy_fork(ip, XFS_ATTR_FORK);
+
+	if (ip->i_itemp) {
+		ASSERT(!(ip->i_itemp->ili_item.li_flags & XFS_LI_IN_AIL));
+		xfs_inode_item_destroy(ip);
+		ip->i_itemp = NULL;
+	}
+
+	/* asserts to verify all state is correct here */
+	ASSERT(atomic_read(&ip->i_pincount) == 0);
+	ASSERT(!spin_is_locked(&ip->i_flags_lock));
+	ASSERT(!xfs_isiflocked(ip));
+
+	/*
+	 * Because we use RCU freeing we need to ensure the inode always
+	 * appears to be reclaimed with an invalid inode number when in the
+	 * free state. The ip->i_flags_lock provides the barrier against lookup
+	 * races.
+	 */
+	spin_lock(&ip->i_flags_lock);
+	ip->i_flags = XFS_IRECLAIM;
+	ip->i_ino = 0;
+	spin_unlock(&ip->i_flags_lock);
+
+	call_rcu(&VFS_I(ip)->i_rcu, xfs_inode_free_callback);
+}
+
+/*
+ * Check the validity of the inode we just found it the cache
+ */
+static int
+xfs_iget_cache_hit(
+	struct xfs_perag	*pag,
+	struct xfs_inode	*ip,
+	xfs_ino_t		ino,
+	int			flags,
+	int			lock_flags) __releases(RCU)
+{
+	struct inode		*inode = VFS_I(ip);
+	struct xfs_mount	*mp = ip->i_mount;
+	int			error;
+
+	/*
+	 * check for re-use of an inode within an RCU grace period due to the
+	 * radix tree nodes not being updated yet. We monitor for this by
+	 * setting the inode number to zero before freeing the inode structure.
+	 * If the inode has been reallocated and set up, then the inode number
+	 * will not match, so check for that, too.
+	 */
+	spin_lock(&ip->i_flags_lock);
+	if (ip->i_ino != ino) {
+		trace_xfs_iget_skip(ip);
+		XFS_STATS_INC(xs_ig_frecycle);
+		error = EAGAIN;
+		goto out_error;
+	}
+
+
+	/*
+	 * If we are racing with another cache hit that is currently
+	 * instantiating this inode or currently recycling it out of
+	 * reclaimabe state, wait for the initialisation to complete
+	 * before continuing.
+	 *
+	 * XXX(hch): eventually we should do something equivalent to
+	 *	     wait_on_inode to wait for these flags to be cleared
+	 *	     instead of polling for it.
+	 */
+	if (ip->i_flags & (XFS_INEW|XFS_IRECLAIM)) {
+		trace_xfs_iget_skip(ip);
+		XFS_STATS_INC(xs_ig_frecycle);
+		error = EAGAIN;
+		goto out_error;
+	}
+
+	/*
+	 * If lookup is racing with unlink return an error immediately.
+	 */
+	if (ip->i_d.di_mode == 0 && !(flags & XFS_IGET_CREATE)) {
+		error = ENOENT;
+		goto out_error;
+	}
+
+	/*
+	 * If IRECLAIMABLE is set, we've torn down the VFS inode already.
+	 * Need to carefully get it back into useable state.
+	 */
+	if (ip->i_flags & XFS_IRECLAIMABLE) {
+		trace_xfs_iget_reclaim(ip);
+
+		/*
+		 * We need to set XFS_IRECLAIM to prevent xfs_reclaim_inode
+		 * from stomping over us while we recycle the inode.  We can't
+		 * clear the radix tree reclaimable tag yet as it requires
+		 * pag_ici_lock to be held exclusive.
+		 */
+		ip->i_flags |= XFS_IRECLAIM;
+
+		spin_unlock(&ip->i_flags_lock);
+		rcu_read_unlock();
+
+		error = -inode_init_always(mp->m_super, inode);
+		if (error) {
+			/*
+			 * Re-initializing the inode failed, and we are in deep
+			 * trouble.  Try to re-add it to the reclaim list.
+			 */
+			rcu_read_lock();
+			spin_lock(&ip->i_flags_lock);
+
+			ip->i_flags &= ~(XFS_INEW | XFS_IRECLAIM);
+			ASSERT(ip->i_flags & XFS_IRECLAIMABLE);
+			trace_xfs_iget_reclaim_fail(ip);
+			goto out_error;
+		}
+
+		spin_lock(&pag->pag_ici_lock);
+		spin_lock(&ip->i_flags_lock);
+
+		/*
+		 * Clear the per-lifetime state in the inode as we are now
+		 * effectively a new inode and need to return to the initial
+		 * state before reuse occurs.
+		 */
+		ip->i_flags &= ~XFS_IRECLAIM_RESET_FLAGS;
+		ip->i_flags |= XFS_INEW;
+		__xfs_inode_clear_reclaim_tag(mp, pag, ip);
+		inode->i_state = I_NEW;
+
+		ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
+		mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
+
+		spin_unlock(&ip->i_flags_lock);
+		spin_unlock(&pag->pag_ici_lock);
+	} else {
+		/* If the VFS inode is being torn down, pause and try again. */
+		if (!igrab(inode)) {
+			trace_xfs_iget_skip(ip);
+			error = EAGAIN;
+			goto out_error;
+		}
+
+		/* We've got a live one. */
+		spin_unlock(&ip->i_flags_lock);
+		rcu_read_unlock();
+		trace_xfs_iget_hit(ip);
+	}
+
+	if (lock_flags != 0)
+		xfs_ilock(ip, lock_flags);
+
+	xfs_iflags_clear(ip, XFS_ISTALE | XFS_IDONTCACHE);
+	XFS_STATS_INC(xs_ig_found);
+
+	return 0;
+
+out_error:
+	spin_unlock(&ip->i_flags_lock);
+	rcu_read_unlock();
+	return error;
+}
+
+
+static int
+xfs_iget_cache_miss(
+	struct xfs_mount	*mp,
+	struct xfs_perag	*pag,
+	xfs_trans_t		*tp,
+	xfs_ino_t		ino,
+	struct xfs_inode	**ipp,
+	int			flags,
+	int			lock_flags)
+{
+	struct xfs_inode	*ip;
+	int			error;
+	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, ino);
+	int			iflags;
+
+	ip = xfs_inode_alloc(mp, ino);
+	if (!ip)
+		return ENOMEM;
+
+	error = xfs_iread(mp, tp, ip, flags);
+	if (error)
+		goto out_destroy;
+
+	trace_xfs_iget_miss(ip);
+
+	if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) {
+		error = ENOENT;
+		goto out_destroy;
+	}
+
+	/*
+	 * Preload the radix tree so we can insert safely under the
+	 * write spinlock. Note that we cannot sleep inside the preload
+	 * region. Since we can be called from transaction context, don't
+	 * recurse into the file system.
+	 */
+	if (radix_tree_preload(GFP_NOFS)) {
+		error = EAGAIN;
+		goto out_destroy;
+	}
+
+	/*
+	 * Because the inode hasn't been added to the radix-tree yet it can't
+	 * be found by another thread, so we can do the non-sleeping lock here.
+	 */
+	if (lock_flags) {
+		if (!xfs_ilock_nowait(ip, lock_flags))
+			BUG();
+	}
+
+	/*
+	 * These values must be set before inserting the inode into the radix
+	 * tree as the moment it is inserted a concurrent lookup (allowed by the
+	 * RCU locking mechanism) can find it and that lookup must see that this
+	 * is an inode currently under construction (i.e. that XFS_INEW is set).
+	 * The ip->i_flags_lock that protects the XFS_INEW flag forms the
+	 * memory barrier that ensures this detection works correctly at lookup
+	 * time.
+	 */
+	iflags = XFS_INEW;
+	if (flags & XFS_IGET_DONTCACHE)
+		iflags |= XFS_IDONTCACHE;
+	ip->i_udquot = ip->i_gdquot = NULL;
+	xfs_iflags_set(ip, iflags);
+
+	/* insert the new inode */
+	spin_lock(&pag->pag_ici_lock);
+	error = radix_tree_insert(&pag->pag_ici_root, agino, ip);
+	if (unlikely(error)) {
+		WARN_ON(error != -EEXIST);
+		XFS_STATS_INC(xs_ig_dup);
+		error = EAGAIN;
+		goto out_preload_end;
+	}
+	spin_unlock(&pag->pag_ici_lock);
+	radix_tree_preload_end();
+
+	*ipp = ip;
+	return 0;
+
+out_preload_end:
+	spin_unlock(&pag->pag_ici_lock);
+	radix_tree_preload_end();
+	if (lock_flags)
+		xfs_iunlock(ip, lock_flags);
+out_destroy:
+	__destroy_inode(VFS_I(ip));
+	xfs_inode_free(ip);
+	return error;
+}
+
+/*
+ * Look up an inode by number in the given file system.
+ * The inode is looked up in the cache held in each AG.
+ * If the inode is found in the cache, initialise the vfs inode
+ * if necessary.
+ *
+ * If it is not in core, read it in from the file system's device,
+ * add it to the cache and initialise the vfs inode.
+ *
+ * The inode is locked according to the value of the lock_flags parameter.
+ * This flag parameter indicates how and if the inode's IO lock and inode lock
+ * should be taken.
+ *
+ * mp -- the mount point structure for the current file system.  It points
+ *       to the inode hash table.
+ * tp -- a pointer to the current transaction if there is one.  This is
+ *       simply passed through to the xfs_iread() call.
+ * ino -- the number of the inode desired.  This is the unique identifier
+ *        within the file system for the inode being requested.
+ * lock_flags -- flags indicating how to lock the inode.  See the comment
+ *		 for xfs_ilock() for a list of valid values.
+ */
+int
+xfs_iget(
+	xfs_mount_t	*mp,
+	xfs_trans_t	*tp,
+	xfs_ino_t	ino,
+	uint		flags,
+	uint		lock_flags,
+	xfs_inode_t	**ipp)
+{
+	xfs_inode_t	*ip;
+	int		error;
+	xfs_perag_t	*pag;
+	xfs_agino_t	agino;
+
+	/*
+	 * xfs_reclaim_inode() uses the ILOCK to ensure an inode
+	 * doesn't get freed while it's being referenced during a
+	 * radix tree traversal here.  It assumes this function
+	 * aqcuires only the ILOCK (and therefore it has no need to
+	 * involve the IOLOCK in this synchronization).
+	 */
+	ASSERT((lock_flags & (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED)) == 0);
+
+	/* reject inode numbers outside existing AGs */
+	if (!ino || XFS_INO_TO_AGNO(mp, ino) >= mp->m_sb.sb_agcount)
+		return EINVAL;
+
+	/* get the perag structure and ensure that it's inode capable */
+	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ino));
+	agino = XFS_INO_TO_AGINO(mp, ino);
+
+again:
+	error = 0;
+	rcu_read_lock();
+	ip = radix_tree_lookup(&pag->pag_ici_root, agino);
+
+	if (ip) {
+		error = xfs_iget_cache_hit(pag, ip, ino, flags, lock_flags);
+		if (error)
+			goto out_error_or_again;
+	} else {
+		rcu_read_unlock();
+		XFS_STATS_INC(xs_ig_missed);
+
+		error = xfs_iget_cache_miss(mp, pag, tp, ino, &ip,
+							flags, lock_flags);
+		if (error)
+			goto out_error_or_again;
+	}
+	xfs_perag_put(pag);
+
+	*ipp = ip;
+
+	/*
+	 * If we have a real type for an on-disk inode, we can set ops(&unlock)
+	 * now.	 If it's a new inode being created, xfs_ialloc will handle it.
+	 */
+	if (xfs_iflags_test(ip, XFS_INEW) && ip->i_d.di_mode != 0)
+		xfs_setup_inode(ip);
+	return 0;
+
+out_error_or_again:
+	if (error == EAGAIN) {
+		delay(1);
+		goto again;
+	}
+	xfs_perag_put(pag);
+	return error;
+}
 
 /*
  * The inode lookup is done in batches to keep the amount of lock traffic and
@@ -101,8 +516,11 @@ xfs_inode_ag_walk(
 	struct xfs_mount	*mp,
 	struct xfs_perag	*pag,
 	int			(*execute)(struct xfs_inode *ip,
-					   struct xfs_perag *pag, int flags),
-	int			flags)
+					   struct xfs_perag *pag, int flags,
+					   void *args),
+	int			flags,
+	void			*args,
+	int			tag)
 {
 	uint32_t		first_index;
 	int			last_error = 0;
@@ -121,9 +539,17 @@ restart:
 		int		i;
 
 		rcu_read_lock();
-		nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
+
+		if (tag == -1)
+			nr_found = radix_tree_gang_lookup(&pag->pag_ici_root,
 					(void **)batch, first_index,
 					XFS_LOOKUP_BATCH);
+		else
+			nr_found = radix_tree_gang_lookup_tag(
+					&pag->pag_ici_root,
+					(void **) batch, first_index,
+					XFS_LOOKUP_BATCH, tag);
+
 		if (!nr_found) {
 			rcu_read_unlock();
 			break;
@@ -164,7 +590,7 @@ restart:
 		for (i = 0; i < nr_found; i++) {
 			if (!batch[i])
 				continue;
-			error = execute(batch[i], pag, flags);
+			error = execute(batch[i], pag, flags, args);
 			IRELE(batch[i]);
 			if (error == EAGAIN) {
 				skipped++;
@@ -189,12 +615,40 @@ restart:
 	return last_error;
 }
 
+/*
+ * Background scanning to trim post-EOF preallocated space. This is queued
+ * based on the 'background_prealloc_discard_period' tunable (5m by default).
+ */
+STATIC void
+xfs_queue_eofblocks(
+	struct xfs_mount *mp)
+{
+	rcu_read_lock();
+	if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_EOFBLOCKS_TAG))
+		queue_delayed_work(mp->m_eofblocks_workqueue,
+				   &mp->m_eofblocks_work,
+				   msecs_to_jiffies(xfs_eofb_secs * 1000));
+	rcu_read_unlock();
+}
+
+void
+xfs_eofblocks_worker(
+	struct work_struct *work)
+{
+	struct xfs_mount *mp = container_of(to_delayed_work(work),
+				struct xfs_mount, m_eofblocks_work);
+	xfs_icache_free_eofblocks(mp, NULL);
+	xfs_queue_eofblocks(mp);
+}
+
 int
 xfs_inode_ag_iterator(
 	struct xfs_mount	*mp,
 	int			(*execute)(struct xfs_inode *ip,
-					   struct xfs_perag *pag, int flags),
-	int			flags)
+					   struct xfs_perag *pag, int flags,
+					   void *args),
+	int			flags,
+	void			*args)
 {
 	struct xfs_perag	*pag;
 	int			error = 0;
@@ -204,7 +658,7 @@ xfs_inode_ag_iterator(
 	ag = 0;
 	while ((pag = xfs_perag_get(mp, ag))) {
 		ag = pag->pag_agno + 1;
-		error = xfs_inode_ag_walk(mp, pag, execute, flags);
+		error = xfs_inode_ag_walk(mp, pag, execute, flags, args, -1);
 		xfs_perag_put(pag);
 		if (error) {
 			last_error = error;
@@ -215,224 +669,50 @@ xfs_inode_ag_iterator(
 	return XFS_ERROR(last_error);
 }
 
-STATIC int
-xfs_sync_inode_data(
-	struct xfs_inode	*ip,
-	struct xfs_perag	*pag,
-	int			flags)
-{
-	struct inode		*inode = VFS_I(ip);
-	struct address_space *mapping = inode->i_mapping;
-	int			error = 0;
-
-	if (!mapping_tagged(mapping, PAGECACHE_TAG_DIRTY))
-		return 0;
-
-	if (!xfs_ilock_nowait(ip, XFS_IOLOCK_SHARED)) {
-		if (flags & SYNC_TRYLOCK)
-			return 0;
-		xfs_ilock(ip, XFS_IOLOCK_SHARED);
-	}
-
-	error = xfs_flush_pages(ip, 0, -1, (flags & SYNC_WAIT) ?
-				0 : XBF_ASYNC, FI_NONE);
-	xfs_iunlock(ip, XFS_IOLOCK_SHARED);
-	return error;
-}
-
-/*
- * Write out pagecache data for the whole filesystem.
- */
-STATIC int
-xfs_sync_data(
-	struct xfs_mount	*mp,
-	int			flags)
-{
-	int			error;
-
-	ASSERT((flags & ~(SYNC_TRYLOCK|SYNC_WAIT)) == 0);
-
-	error = xfs_inode_ag_iterator(mp, xfs_sync_inode_data, flags);
-	if (error)
-		return XFS_ERROR(error);
-
-	xfs_log_force(mp, (flags & SYNC_WAIT) ? XFS_LOG_SYNC : 0);
-	return 0;
-}
-
-STATIC int
-xfs_sync_fsdata(
-	struct xfs_mount	*mp)
-{
-	struct xfs_buf		*bp;
-	int			error;
-
-	/*
-	 * If the buffer is pinned then push on the log so we won't get stuck
-	 * waiting in the write for someone, maybe ourselves, to flush the log.
-	 *
-	 * Even though we just pushed the log above, we did not have the
-	 * superblock buffer locked at that point so it can become pinned in
-	 * between there and here.
-	 */
-	bp = xfs_getsb(mp, 0);
-	if (xfs_buf_ispinned(bp))
-		xfs_log_force(mp, 0);
-	error = xfs_bwrite(bp);
-	xfs_buf_relse(bp);
-	return error;
-}
-
-/*
- * When remounting a filesystem read-only or freezing the filesystem, we have
- * two phases to execute. This first phase is syncing the data before we
- * quiesce the filesystem, and the second is flushing all the inodes out after
- * we've waited for all the transactions created by the first phase to
- * complete. The second phase ensures that the inodes are written to their
- * location on disk rather than just existing in transactions in the log. This
- * means after a quiesce there is no log replay required to write the inodes to
- * disk (this is the main difference between a sync and a quiesce).
- */
-/*
- * First stage of freeze - no writers will make progress now we are here,
- * so we flush delwri and delalloc buffers here, then wait for all I/O to
- * complete.  Data is frozen at that point. Metadata is not frozen,
- * transactions can still occur here so don't bother emptying the AIL
- * because it'll just get dirty again.
- */
 int
-xfs_quiesce_data(
-	struct xfs_mount	*mp)
-{
-	int			error, error2 = 0;
-
-	/* force out the log */
-	xfs_log_force(mp, XFS_LOG_SYNC);
-
-	/* write superblock and hoover up shutdown errors */
-	error = xfs_sync_fsdata(mp);
-
-	/* mark the log as covered if needed */
-	if (xfs_log_need_covered(mp))
-		error2 = xfs_fs_log_dummy(mp);
-
-	return error ? error : error2;
-}
-
-/*
- * Second stage of a quiesce. The data is already synced, now we have to take
- * care of the metadata. New transactions are already blocked, so we need to
- * wait for any remaining transactions to drain out before proceeding.
- */
-void
-xfs_quiesce_attr(
-	struct xfs_mount	*mp)
-{
-	int	error = 0;
-
-	/* wait for all modifications to complete */
-	while (atomic_read(&mp->m_active_trans) > 0)
-		delay(100);
-
-	/* reclaim inodes to do any IO before the freeze completes */
-	xfs_reclaim_inodes(mp, 0);
-	xfs_reclaim_inodes(mp, SYNC_WAIT);
-
-	/* flush all pending changes from the AIL */
-	xfs_ail_push_all_sync(mp->m_ail);
-
-	/*
-	 * Just warn here till VFS can correctly support
-	 * read-only remount without racing.
-	 */
-	WARN_ON(atomic_read(&mp->m_active_trans) != 0);
-
-	/* Push the superblock and write an unmount record */
-	error = xfs_log_sbcount(mp);
-	if (error)
-		xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
-				"Frozen image may not be consistent.");
-	xfs_log_unmount_write(mp);
-
-	/*
-	 * At this point we might have modified the superblock again and thus
-	 * added an item to the AIL, thus flush it again.
-	 */
-	xfs_ail_push_all_sync(mp->m_ail);
-
-	/*
-	 * The superblock buffer is uncached and xfsaild_push() will lock and
-	 * set the XBF_ASYNC flag on the buffer. We cannot do xfs_buf_iowait()
-	 * here but a lock on the superblock buffer will block until iodone()
-	 * has completed.
-	 */
-	xfs_buf_lock(mp->m_sb_bp);
-	xfs_buf_unlock(mp->m_sb_bp);
-}
-
-static void
-xfs_syncd_queue_sync(
-	struct xfs_mount        *mp)
-{
-	queue_delayed_work(xfs_syncd_wq, &mp->m_sync_work,
-				msecs_to_jiffies(xfs_syncd_centisecs * 10));
-}
-
-/*
- * Every sync period we need to unpin all items, reclaim inodes and sync
- * disk quotas.  We might need to cover the log to indicate that the
- * filesystem is idle and not frozen.
- */
-STATIC void
-xfs_sync_worker(
-	struct work_struct *work)
+xfs_inode_ag_iterator_tag(
+	struct xfs_mount	*mp,
+	int			(*execute)(struct xfs_inode *ip,
+					   struct xfs_perag *pag, int flags,
+					   void *args),
+	int			flags,
+	void			*args,
+	int			tag)
 {
-	struct xfs_mount *mp = container_of(to_delayed_work(work),
-					struct xfs_mount, m_sync_work);
-	int		error;
-
-	/*
-	 * We shouldn't write/force the log if we are in the mount/unmount
-	 * process or on a read only filesystem. The workqueue still needs to be
-	 * active in both cases, however, because it is used for inode reclaim
-	 * during these times.  Use the MS_ACTIVE flag to avoid doing anything
-	 * during mount.  Doing work during unmount is avoided by calling
-	 * cancel_delayed_work_sync on this work queue before tearing down
-	 * the ail and the log in xfs_log_unmount.
-	 */
-	if (!(mp->m_super->s_flags & MS_ACTIVE) &&
-	    !(mp->m_flags & XFS_MOUNT_RDONLY)) {
-		/* dgc: errors ignored here */
-		if (mp->m_super->s_writers.frozen == SB_UNFROZEN &&
-		    xfs_log_need_covered(mp))
-			error = xfs_fs_log_dummy(mp);
-		else
-			xfs_log_force(mp, 0);
+	struct xfs_perag	*pag;
+	int			error = 0;
+	int			last_error = 0;
+	xfs_agnumber_t		ag;
 
-		/* start pushing all the metadata that is currently
-		 * dirty */
-		xfs_ail_push_all(mp->m_ail);
+	ag = 0;
+	while ((pag = xfs_perag_get_tag(mp, ag, tag))) {
+		ag = pag->pag_agno + 1;
+		error = xfs_inode_ag_walk(mp, pag, execute, flags, args, tag);
+		xfs_perag_put(pag);
+		if (error) {
+			last_error = error;
+			if (error == EFSCORRUPTED)
+				break;
+		}
 	}
-
-	/* queue us up again */
-	xfs_syncd_queue_sync(mp);
+	return XFS_ERROR(last_error);
 }
 
 /*
  * Queue a new inode reclaim pass if there are reclaimable inodes and there
  * isn't a reclaim pass already in progress. By default it runs every 5s based
- * on the xfs syncd work default of 30s. Perhaps this should have it's own
+ * on the xfs periodic sync default of 30s. Perhaps this should have it's own
  * tunable, but that can be done if this method proves to be ineffective or too
  * aggressive.
  */
 static void
-xfs_syncd_queue_reclaim(
+xfs_reclaim_work_queue(
 	struct xfs_mount        *mp)
 {
 
 	rcu_read_lock();
 	if (radix_tree_tagged(&mp->m_perag_tree, XFS_ICI_RECLAIM_TAG)) {
-		queue_delayed_work(xfs_syncd_wq, &mp->m_reclaim_work,
+		queue_delayed_work(mp->m_reclaim_workqueue, &mp->m_reclaim_work,
 			msecs_to_jiffies(xfs_syncd_centisecs / 6 * 10));
 	}
 	rcu_read_unlock();
@@ -445,7 +725,7 @@ xfs_syncd_queue_reclaim(
  * goes low. It scans as quickly as possible avoiding locked inodes or those
  * already being flushed, and once done schedules a future pass.
  */
-STATIC void
+void
 xfs_reclaim_worker(
 	struct work_struct *work)
 {
@@ -453,65 +733,10 @@ xfs_reclaim_worker(
 					struct xfs_mount, m_reclaim_work);
 
 	xfs_reclaim_inodes(mp, SYNC_TRYLOCK);
-	xfs_syncd_queue_reclaim(mp);
+	xfs_reclaim_work_queue(mp);
 }
 
-/*
- * Flush delayed allocate data, attempting to free up reserved space
- * from existing allocations.  At this point a new allocation attempt
- * has failed with ENOSPC and we are in the process of scratching our
- * heads, looking about for more room.
- *
- * Queue a new data flush if there isn't one already in progress and
- * wait for completion of the flush. This means that we only ever have one
- * inode flush in progress no matter how many ENOSPC events are occurring and
- * so will prevent the system from bogging down due to every concurrent
- * ENOSPC event scanning all the active inodes in the system for writeback.
- */
-void
-xfs_flush_inodes(
-	struct xfs_inode	*ip)
-{
-	struct xfs_mount	*mp = ip->i_mount;
-
-	queue_work(xfs_syncd_wq, &mp->m_flush_work);
-	flush_work(&mp->m_flush_work);
-}
-
-STATIC void
-xfs_flush_worker(
-	struct work_struct *work)
-{
-	struct xfs_mount *mp = container_of(work,
-					struct xfs_mount, m_flush_work);
-
-	xfs_sync_data(mp, SYNC_TRYLOCK);
-	xfs_sync_data(mp, SYNC_TRYLOCK | SYNC_WAIT);
-}
-
-int
-xfs_syncd_init(
-	struct xfs_mount	*mp)
-{
-	INIT_WORK(&mp->m_flush_work, xfs_flush_worker);
-	INIT_DELAYED_WORK(&mp->m_sync_work, xfs_sync_worker);
-	INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
-
-	xfs_syncd_queue_sync(mp);
-
-	return 0;
-}
-
-void
-xfs_syncd_stop(
-	struct xfs_mount	*mp)
-{
-	cancel_delayed_work_sync(&mp->m_sync_work);
-	cancel_delayed_work_sync(&mp->m_reclaim_work);
-	cancel_work_sync(&mp->m_flush_work);
-}
-
-void
+static void
 __xfs_inode_set_reclaim_tag(
 	struct xfs_perag	*pag,
 	struct xfs_inode	*ip)
@@ -529,7 +754,7 @@ __xfs_inode_set_reclaim_tag(
 		spin_unlock(&ip->i_mount->m_perag_lock);
 
 		/* schedule periodic background inode reclaim */
-		xfs_syncd_queue_reclaim(ip->i_mount);
+		xfs_reclaim_work_queue(ip->i_mount);
 
 		trace_xfs_perag_set_reclaim(ip->i_mount, pag->pag_agno,
 							-1, _RET_IP_);
@@ -577,7 +802,7 @@ __xfs_inode_clear_reclaim(
 	}
 }
 
-void
+STATIC void
 __xfs_inode_clear_reclaim_tag(
 	xfs_mount_t	*mp,
 	xfs_perag_t	*pag,
@@ -787,9 +1012,9 @@ out:
 	/*
 	 * We could return EAGAIN here to make reclaim rescan the inode tree in
 	 * a short while. However, this just burns CPU time scanning the tree
-	 * waiting for IO to complete and xfssyncd never goes back to the idle
-	 * state. Instead, return 0 to let the next scheduled background reclaim
-	 * attempt to reclaim the inode again.
+	 * waiting for IO to complete and the reclaim work never goes back to
+	 * the idle state. Instead, return 0 to let the next scheduled
+	 * background reclaim attempt to reclaim the inode again.
 	 */
 	return 0;
 }
@@ -800,7 +1025,7 @@ out:
  * then a shut down during filesystem unmount reclaim walk leak all the
  * unreclaimed inodes.
  */
-int
+STATIC int
 xfs_reclaim_inodes_ag(
 	struct xfs_mount	*mp,
 	int			flags,
@@ -945,7 +1170,7 @@ xfs_reclaim_inodes_nr(
 	int			nr_to_scan)
 {
 	/* kick background reclaimer and push the AIL */
-	xfs_syncd_queue_reclaim(mp);
+	xfs_reclaim_work_queue(mp);
 	xfs_ail_push_all(mp->m_ail);
 
 	xfs_reclaim_inodes_ag(mp, SYNC_TRYLOCK | SYNC_WAIT, &nr_to_scan);
@@ -971,3 +1196,146 @@ xfs_reclaim_inodes_count(
 	return reclaimable;
 }
 
+STATIC int
+xfs_inode_match_id(
+	struct xfs_inode	*ip,
+	struct xfs_eofblocks	*eofb)
+{
+	if (eofb->eof_flags & XFS_EOF_FLAGS_UID &&
+	    ip->i_d.di_uid != eofb->eof_uid)
+		return 0;
+
+	if (eofb->eof_flags & XFS_EOF_FLAGS_GID &&
+	    ip->i_d.di_gid != eofb->eof_gid)
+		return 0;
+
+	if (eofb->eof_flags & XFS_EOF_FLAGS_PRID &&
+	    xfs_get_projid(ip) != eofb->eof_prid)
+		return 0;
+
+	return 1;
+}
+
+STATIC int
+xfs_inode_free_eofblocks(
+	struct xfs_inode	*ip,
+	struct xfs_perag	*pag,
+	int			flags,
+	void			*args)
+{
+	int ret;
+	struct xfs_eofblocks *eofb = args;
+
+	if (!xfs_can_free_eofblocks(ip, false)) {
+		/* inode could be preallocated or append-only */
+		trace_xfs_inode_free_eofblocks_invalid(ip);
+		xfs_inode_clear_eofblocks_tag(ip);
+		return 0;
+	}
+
+	/*
+	 * If the mapping is dirty the operation can block and wait for some
+	 * time. Unless we are waiting, skip it.
+	 */
+	if (!(flags & SYNC_WAIT) &&
+	    mapping_tagged(VFS_I(ip)->i_mapping, PAGECACHE_TAG_DIRTY))
+		return 0;
+
+	if (eofb) {
+		if (!xfs_inode_match_id(ip, eofb))
+			return 0;
+
+		/* skip the inode if the file size is too small */
+		if (eofb->eof_flags & XFS_EOF_FLAGS_MINFILESIZE &&
+		    XFS_ISIZE(ip) < eofb->eof_min_file_size)
+			return 0;
+	}
+
+	ret = xfs_free_eofblocks(ip->i_mount, ip, true);
+
+	/* don't revisit the inode if we're not waiting */
+	if (ret == EAGAIN && !(flags & SYNC_WAIT))
+		ret = 0;
+
+	return ret;
+}
+
+int
+xfs_icache_free_eofblocks(
+	struct xfs_mount	*mp,
+	struct xfs_eofblocks	*eofb)
+{
+	int flags = SYNC_TRYLOCK;
+
+	if (eofb && (eofb->eof_flags & XFS_EOF_FLAGS_SYNC))
+		flags = SYNC_WAIT;
+
+	return xfs_inode_ag_iterator_tag(mp, xfs_inode_free_eofblocks, flags,
+					 eofb, XFS_ICI_EOFBLOCKS_TAG);
+}
+
+void
+xfs_inode_set_eofblocks_tag(
+	xfs_inode_t	*ip)
+{
+	struct xfs_mount *mp = ip->i_mount;
+	struct xfs_perag *pag;
+	int tagged;
+
+	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
+	spin_lock(&pag->pag_ici_lock);
+	trace_xfs_inode_set_eofblocks_tag(ip);
+
+	tagged = radix_tree_tagged(&pag->pag_ici_root,
+				   XFS_ICI_EOFBLOCKS_TAG);
+	radix_tree_tag_set(&pag->pag_ici_root,
+			   XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino),
+			   XFS_ICI_EOFBLOCKS_TAG);
+	if (!tagged) {
+		/* propagate the eofblocks tag up into the perag radix tree */
+		spin_lock(&ip->i_mount->m_perag_lock);
+		radix_tree_tag_set(&ip->i_mount->m_perag_tree,
+				   XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
+				   XFS_ICI_EOFBLOCKS_TAG);
+		spin_unlock(&ip->i_mount->m_perag_lock);
+
+		/* kick off background trimming */
+		xfs_queue_eofblocks(ip->i_mount);
+
+		trace_xfs_perag_set_eofblocks(ip->i_mount, pag->pag_agno,
+					      -1, _RET_IP_);
+	}
+
+	spin_unlock(&pag->pag_ici_lock);
+	xfs_perag_put(pag);
+}
+
+void
+xfs_inode_clear_eofblocks_tag(
+	xfs_inode_t	*ip)
+{
+	struct xfs_mount *mp = ip->i_mount;
+	struct xfs_perag *pag;
+
+	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ip->i_ino));
+	spin_lock(&pag->pag_ici_lock);
+	trace_xfs_inode_clear_eofblocks_tag(ip);
+
+	radix_tree_tag_clear(&pag->pag_ici_root,
+			     XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino),
+			     XFS_ICI_EOFBLOCKS_TAG);
+	if (!radix_tree_tagged(&pag->pag_ici_root, XFS_ICI_EOFBLOCKS_TAG)) {
+		/* clear the eofblocks tag from the perag radix tree */
+		spin_lock(&ip->i_mount->m_perag_lock);
+		radix_tree_tag_clear(&ip->i_mount->m_perag_tree,
+				     XFS_INO_TO_AGNO(ip->i_mount, ip->i_ino),
+				     XFS_ICI_EOFBLOCKS_TAG);
+		spin_unlock(&ip->i_mount->m_perag_lock);
+		trace_xfs_perag_clear_eofblocks(ip->i_mount, pag->pag_agno,
+					       -1, _RET_IP_);
+	}
+
+	spin_unlock(&pag->pag_ici_lock);
+	xfs_perag_put(pag);
+}
+

fs/xfs/xfs_sync.h → fs/xfs/xfs_icache.h

@@ -24,28 +24,30 @@ struct xfs_perag;
 #define SYNC_WAIT		0x0001	/* wait for i/o to complete */
 #define SYNC_TRYLOCK		0x0002  /* only try to lock inodes */
 
-extern struct workqueue_struct	*xfs_syncd_wq;	/* sync workqueue */
+int xfs_iget(struct xfs_mount *mp, struct xfs_trans *tp, xfs_ino_t ino,
+	     uint flags, uint lock_flags, xfs_inode_t **ipp);
 
-int xfs_syncd_init(struct xfs_mount *mp);
-void xfs_syncd_stop(struct xfs_mount *mp);
-
-int xfs_quiesce_data(struct xfs_mount *mp);
-void xfs_quiesce_attr(struct xfs_mount *mp);
-
-void xfs_flush_inodes(struct xfs_inode *ip);
+void xfs_reclaim_worker(struct work_struct *work);
 
 int xfs_reclaim_inodes(struct xfs_mount *mp, int mode);
 int xfs_reclaim_inodes_count(struct xfs_mount *mp);
 void xfs_reclaim_inodes_nr(struct xfs_mount *mp, int nr_to_scan);
 
 void xfs_inode_set_reclaim_tag(struct xfs_inode *ip);
-void __xfs_inode_set_reclaim_tag(struct xfs_perag *pag, struct xfs_inode *ip);
-void __xfs_inode_clear_reclaim_tag(struct xfs_mount *mp, struct xfs_perag *pag,
-				struct xfs_inode *ip);
+
+void xfs_inode_set_eofblocks_tag(struct xfs_inode *ip);
+void xfs_inode_clear_eofblocks_tag(struct xfs_inode *ip);
+int xfs_icache_free_eofblocks(struct xfs_mount *, struct xfs_eofblocks *);
+void xfs_eofblocks_worker(struct work_struct *);
 
 int xfs_sync_inode_grab(struct xfs_inode *ip);
 int xfs_inode_ag_iterator(struct xfs_mount *mp,
-	int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag, int flags),
-	int flags);
+	int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag,
+		int flags, void *args),
+	int flags, void *args);
+int xfs_inode_ag_iterator_tag(struct xfs_mount *mp,
+	int (*execute)(struct xfs_inode *ip, struct xfs_perag *pag,
+		int flags, void *args),
+	int flags, void *args, int tag);
 
 #endif

fs/xfs/xfs_iget.c

@@ -1,705 +0,0 @@
-/*
- * Copyright (c) 2000-2005 Silicon Graphics, Inc.
- * All Rights Reserved.
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it would be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write the Free Software Foundation,
- * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
- */
-#include "xfs.h"
-#include "xfs_fs.h"
-#include "xfs_types.h"
-#include "xfs_acl.h"
-#include "xfs_log.h"
-#include "xfs_inum.h"
-#include "xfs_trans.h"
-#include "xfs_sb.h"
-#include "xfs_ag.h"
-#include "xfs_mount.h"
-#include "xfs_bmap_btree.h"
-#include "xfs_alloc_btree.h"
-#include "xfs_ialloc_btree.h"
-#include "xfs_dinode.h"
-#include "xfs_inode.h"
-#include "xfs_btree.h"
-#include "xfs_ialloc.h"
-#include "xfs_quota.h"
-#include "xfs_utils.h"
-#include "xfs_trans_priv.h"
-#include "xfs_inode_item.h"
-#include "xfs_bmap.h"
-#include "xfs_trace.h"
-
-
-/*
- * Allocate and initialise an xfs_inode.
- */
-STATIC struct xfs_inode *
-xfs_inode_alloc(
-	struct xfs_mount	*mp,
-	xfs_ino_t		ino)
-{
-	struct xfs_inode	*ip;
-
-	/*
-	 * if this didn't occur in transactions, we could use
-	 * KM_MAYFAIL and return NULL here on ENOMEM. Set the
-	 * code up to do this anyway.
-	 */
-	ip = kmem_zone_alloc(xfs_inode_zone, KM_SLEEP);
-	if (!ip)
-		return NULL;
-	if (inode_init_always(mp->m_super, VFS_I(ip))) {
-		kmem_zone_free(xfs_inode_zone, ip);
-		return NULL;
-	}
-
-	ASSERT(atomic_read(&ip->i_pincount) == 0);
-	ASSERT(!spin_is_locked(&ip->i_flags_lock));
-	ASSERT(!xfs_isiflocked(ip));
-	ASSERT(ip->i_ino == 0);
-
-	mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
-
-	/* initialise the xfs inode */
-	ip->i_ino = ino;
-	ip->i_mount = mp;
-	memset(&ip->i_imap, 0, sizeof(struct xfs_imap));
-	ip->i_afp = NULL;
-	memset(&ip->i_df, 0, sizeof(xfs_ifork_t));
-	ip->i_flags = 0;
-	ip->i_delayed_blks = 0;
-	memset(&ip->i_d, 0, sizeof(xfs_icdinode_t));
-
-	return ip;
-}
-
-STATIC void
-xfs_inode_free_callback(
-	struct rcu_head		*head)
-{
-	struct inode		*inode = container_of(head, struct inode, i_rcu);
-	struct xfs_inode	*ip = XFS_I(inode);
-
-	kmem_zone_free(xfs_inode_zone, ip);
-}
-
-void
-xfs_inode_free(
-	struct xfs_inode	*ip)
-{
-	switch (ip->i_d.di_mode & S_IFMT) {
-	case S_IFREG:
-	case S_IFDIR:
-	case S_IFLNK:
-		xfs_idestroy_fork(ip, XFS_DATA_FORK);
-		break;
-	}
-
-	if (ip->i_afp)
-		xfs_idestroy_fork(ip, XFS_ATTR_FORK);
-
-	if (ip->i_itemp) {
-		ASSERT(!(ip->i_itemp->ili_item.li_flags & XFS_LI_IN_AIL));
-		xfs_inode_item_destroy(ip);
-		ip->i_itemp = NULL;
-	}
-
-	/* asserts to verify all state is correct here */
-	ASSERT(atomic_read(&ip->i_pincount) == 0);
-	ASSERT(!spin_is_locked(&ip->i_flags_lock));
-	ASSERT(!xfs_isiflocked(ip));
-
-	/*
-	 * Because we use RCU freeing we need to ensure the inode always
-	 * appears to be reclaimed with an invalid inode number when in the
-	 * free state. The ip->i_flags_lock provides the barrier against lookup
-	 * races.
-	 */
-	spin_lock(&ip->i_flags_lock);
-	ip->i_flags = XFS_IRECLAIM;
-	ip->i_ino = 0;
-	spin_unlock(&ip->i_flags_lock);
-
-	call_rcu(&VFS_I(ip)->i_rcu, xfs_inode_free_callback);
-}
-
-/*
- * Check the validity of the inode we just found it the cache
- */
-static int
-xfs_iget_cache_hit(
-	struct xfs_perag	*pag,
-	struct xfs_inode	*ip,
-	xfs_ino_t		ino,
-	int			flags,
-	int			lock_flags) __releases(RCU)
-{
-	struct inode		*inode = VFS_I(ip);
-	struct xfs_mount	*mp = ip->i_mount;
-	int			error;
-
-	/*
-	 * check for re-use of an inode within an RCU grace period due to the
-	 * radix tree nodes not being updated yet. We monitor for this by
-	 * setting the inode number to zero before freeing the inode structure.
-	 * If the inode has been reallocated and set up, then the inode number
-	 * will not match, so check for that, too.
-	 */
-	spin_lock(&ip->i_flags_lock);
-	if (ip->i_ino != ino) {
-		trace_xfs_iget_skip(ip);
-		XFS_STATS_INC(xs_ig_frecycle);
-		error = EAGAIN;
-		goto out_error;
-	}
-
-
-	/*
-	 * If we are racing with another cache hit that is currently
-	 * instantiating this inode or currently recycling it out of
-	 * reclaimabe state, wait for the initialisation to complete
-	 * before continuing.
-	 *
-	 * XXX(hch): eventually we should do something equivalent to
-	 *	     wait_on_inode to wait for these flags to be cleared
-	 *	     instead of polling for it.
-	 */
-	if (ip->i_flags & (XFS_INEW|XFS_IRECLAIM)) {
-		trace_xfs_iget_skip(ip);
-		XFS_STATS_INC(xs_ig_frecycle);
-		error = EAGAIN;
-		goto out_error;
-	}
-
-	/*
-	 * If lookup is racing with unlink return an error immediately.
-	 */
-	if (ip->i_d.di_mode == 0 && !(flags & XFS_IGET_CREATE)) {
-		error = ENOENT;
-		goto out_error;
-	}
-
-	/*
-	 * If IRECLAIMABLE is set, we've torn down the VFS inode already.
-	 * Need to carefully get it back into useable state.
-	 */
-	if (ip->i_flags & XFS_IRECLAIMABLE) {
-		trace_xfs_iget_reclaim(ip);
-
-		/*
-		 * We need to set XFS_IRECLAIM to prevent xfs_reclaim_inode
-		 * from stomping over us while we recycle the inode.  We can't
-		 * clear the radix tree reclaimable tag yet as it requires
-		 * pag_ici_lock to be held exclusive.
-		 */
-		ip->i_flags |= XFS_IRECLAIM;
-
-		spin_unlock(&ip->i_flags_lock);
-		rcu_read_unlock();
-
-		error = -inode_init_always(mp->m_super, inode);
-		if (error) {
-			/*
-			 * Re-initializing the inode failed, and we are in deep
-			 * trouble.  Try to re-add it to the reclaim list.
-			 */
-			rcu_read_lock();
-			spin_lock(&ip->i_flags_lock);
-
-			ip->i_flags &= ~(XFS_INEW | XFS_IRECLAIM);
-			ASSERT(ip->i_flags & XFS_IRECLAIMABLE);
-			trace_xfs_iget_reclaim_fail(ip);
-			goto out_error;
-		}
-
-		spin_lock(&pag->pag_ici_lock);
-		spin_lock(&ip->i_flags_lock);
-
-		/*
-		 * Clear the per-lifetime state in the inode as we are now
-		 * effectively a new inode and need to return to the initial
-		 * state before reuse occurs.
-		 */
-		ip->i_flags &= ~XFS_IRECLAIM_RESET_FLAGS;
-		ip->i_flags |= XFS_INEW;
-		__xfs_inode_clear_reclaim_tag(mp, pag, ip);
-		inode->i_state = I_NEW;
-
-		ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
-		mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
-
-		spin_unlock(&ip->i_flags_lock);
-		spin_unlock(&pag->pag_ici_lock);
-	} else {
-		/* If the VFS inode is being torn down, pause and try again. */
-		if (!igrab(inode)) {
-			trace_xfs_iget_skip(ip);
-			error = EAGAIN;
-			goto out_error;
-		}
-
-		/* We've got a live one. */
-		spin_unlock(&ip->i_flags_lock);
-		rcu_read_unlock();
-		trace_xfs_iget_hit(ip);
-	}
-
-	if (lock_flags != 0)
-		xfs_ilock(ip, lock_flags);
-
-	xfs_iflags_clear(ip, XFS_ISTALE | XFS_IDONTCACHE);
-	XFS_STATS_INC(xs_ig_found);
-
-	return 0;
-
-out_error:
-	spin_unlock(&ip->i_flags_lock);
-	rcu_read_unlock();
-	return error;
-}
-
-
-static int
-xfs_iget_cache_miss(
-	struct xfs_mount	*mp,
-	struct xfs_perag	*pag,
-	xfs_trans_t		*tp,
-	xfs_ino_t		ino,
-	struct xfs_inode	**ipp,
-	int			flags,
-	int			lock_flags)
-{
-	struct xfs_inode	*ip;
-	int			error;
-	xfs_agino_t		agino = XFS_INO_TO_AGINO(mp, ino);
-	int			iflags;
-
-	ip = xfs_inode_alloc(mp, ino);
-	if (!ip)
-		return ENOMEM;
-
-	error = xfs_iread(mp, tp, ip, flags);
-	if (error)
-		goto out_destroy;
-
-	trace_xfs_iget_miss(ip);
-
-	if ((ip->i_d.di_mode == 0) && !(flags & XFS_IGET_CREATE)) {
-		error = ENOENT;
-		goto out_destroy;
-	}
-
-	/*
-	 * Preload the radix tree so we can insert safely under the
-	 * write spinlock. Note that we cannot sleep inside the preload
-	 * region. Since we can be called from transaction context, don't
-	 * recurse into the file system.
-	 */
-	if (radix_tree_preload(GFP_NOFS)) {
-		error = EAGAIN;
-		goto out_destroy;
-	}
-
-	/*
-	 * Because the inode hasn't been added to the radix-tree yet it can't
-	 * be found by another thread, so we can do the non-sleeping lock here.
-	 */
-	if (lock_flags) {
-		if (!xfs_ilock_nowait(ip, lock_flags))
-			BUG();
-	}
-
-	/*
-	 * These values must be set before inserting the inode into the radix
-	 * tree as the moment it is inserted a concurrent lookup (allowed by the
-	 * RCU locking mechanism) can find it and that lookup must see that this
-	 * is an inode currently under construction (i.e. that XFS_INEW is set).
-	 * The ip->i_flags_lock that protects the XFS_INEW flag forms the
-	 * memory barrier that ensures this detection works correctly at lookup
-	 * time.
-	 */
-	iflags = XFS_INEW;
-	if (flags & XFS_IGET_DONTCACHE)
-		iflags |= XFS_IDONTCACHE;
-	ip->i_udquot = ip->i_gdquot = NULL;
-	xfs_iflags_set(ip, iflags);
-
-	/* insert the new inode */
-	spin_lock(&pag->pag_ici_lock);
-	error = radix_tree_insert(&pag->pag_ici_root, agino, ip);
-	if (unlikely(error)) {
-		WARN_ON(error != -EEXIST);
-		XFS_STATS_INC(xs_ig_dup);
-		error = EAGAIN;
-		goto out_preload_end;
-	}
-	spin_unlock(&pag->pag_ici_lock);
-	radix_tree_preload_end();
-
-	*ipp = ip;
-	return 0;
-
-out_preload_end:
-	spin_unlock(&pag->pag_ici_lock);
-	radix_tree_preload_end();
-	if (lock_flags)
-		xfs_iunlock(ip, lock_flags);
-out_destroy:
-	__destroy_inode(VFS_I(ip));
-	xfs_inode_free(ip);
-	return error;
-}
-
-/*
- * Look up an inode by number in the given file system.
- * The inode is looked up in the cache held in each AG.
- * If the inode is found in the cache, initialise the vfs inode
- * if necessary.
- *
- * If it is not in core, read it in from the file system's device,
- * add it to the cache and initialise the vfs inode.
- *
- * The inode is locked according to the value of the lock_flags parameter.
- * This flag parameter indicates how and if the inode's IO lock and inode lock
- * should be taken.
- *
- * mp -- the mount point structure for the current file system.  It points
- *       to the inode hash table.
- * tp -- a pointer to the current transaction if there is one.  This is
- *       simply passed through to the xfs_iread() call.
- * ino -- the number of the inode desired.  This is the unique identifier
- *        within the file system for the inode being requested.
- * lock_flags -- flags indicating how to lock the inode.  See the comment
- *		 for xfs_ilock() for a list of valid values.
- */
-int
-xfs_iget(
-	xfs_mount_t	*mp,
-	xfs_trans_t	*tp,
-	xfs_ino_t	ino,
-	uint		flags,
-	uint		lock_flags,
-	xfs_inode_t	**ipp)
-{
-	xfs_inode_t	*ip;
-	int		error;
-	xfs_perag_t	*pag;
-	xfs_agino_t	agino;
-
-	/*
-	 * xfs_reclaim_inode() uses the ILOCK to ensure an inode
-	 * doesn't get freed while it's being referenced during a
-	 * radix tree traversal here.  It assumes this function
-	 * aqcuires only the ILOCK (and therefore it has no need to
-	 * involve the IOLOCK in this synchronization).
-	 */
-	ASSERT((lock_flags & (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED)) == 0);
-
-	/* reject inode numbers outside existing AGs */
-	if (!ino || XFS_INO_TO_AGNO(mp, ino) >= mp->m_sb.sb_agcount)
-		return EINVAL;
-
-	/* get the perag structure and ensure that it's inode capable */
-	pag = xfs_perag_get(mp, XFS_INO_TO_AGNO(mp, ino));
-	agino = XFS_INO_TO_AGINO(mp, ino);
-
-again:
-	error = 0;
-	rcu_read_lock();
-	ip = radix_tree_lookup(&pag->pag_ici_root, agino);
-
-	if (ip) {
-		error = xfs_iget_cache_hit(pag, ip, ino, flags, lock_flags);
-		if (error)
-			goto out_error_or_again;
-	} else {
-		rcu_read_unlock();
-		XFS_STATS_INC(xs_ig_missed);
-
-		error = xfs_iget_cache_miss(mp, pag, tp, ino, &ip,
-							flags, lock_flags);
-		if (error)
-			goto out_error_or_again;
-	}
-	xfs_perag_put(pag);
-
-	*ipp = ip;
-
-	/*
-	 * If we have a real type for an on-disk inode, we can set ops(&unlock)
-	 * now.	 If it's a new inode being created, xfs_ialloc will handle it.
-	 */
-	if (xfs_iflags_test(ip, XFS_INEW) && ip->i_d.di_mode != 0)
-		xfs_setup_inode(ip);
-	return 0;
-
-out_error_or_again:
-	if (error == EAGAIN) {
-		delay(1);
-		goto again;
-	}
-	xfs_perag_put(pag);
-	return error;
-}
-
-/*
- * This is a wrapper routine around the xfs_ilock() routine
- * used to centralize some grungy code.  It is used in places
- * that wish to lock the inode solely for reading the extents.
- * The reason these places can't just call xfs_ilock(SHARED)
- * is that the inode lock also guards to bringing in of the
- * extents from disk for a file in b-tree format.  If the inode
- * is in b-tree format, then we need to lock the inode exclusively
- * until the extents are read in.  Locking it exclusively all
- * the time would limit our parallelism unnecessarily, though.
- * What we do instead is check to see if the extents have been
- * read in yet, and only lock the inode exclusively if they
- * have not.
- *
- * The function returns a value which should be given to the
- * corresponding xfs_iunlock_map_shared().  This value is
- * the mode in which the lock was actually taken.
- */
-uint
-xfs_ilock_map_shared(
-	xfs_inode_t	*ip)
-{
-	uint	lock_mode;
-
-	if ((ip->i_d.di_format == XFS_DINODE_FMT_BTREE) &&
-	    ((ip->i_df.if_flags & XFS_IFEXTENTS) == 0)) {
-		lock_mode = XFS_ILOCK_EXCL;
-	} else {
-		lock_mode = XFS_ILOCK_SHARED;
-	}
-
-	xfs_ilock(ip, lock_mode);
-
-	return lock_mode;
-}
-
-/*
- * This is simply the unlock routine to go with xfs_ilock_map_shared().
- * All it does is call xfs_iunlock() with the given lock_mode.
- */
-void
-xfs_iunlock_map_shared(
-	xfs_inode_t	*ip,
-	unsigned int	lock_mode)
-{
-	xfs_iunlock(ip, lock_mode);
-}
-
-/*
- * The xfs inode contains 2 locks: a multi-reader lock called the
- * i_iolock and a multi-reader lock called the i_lock.  This routine
- * allows either or both of the locks to be obtained.
- *
- * The 2 locks should always be ordered so that the IO lock is
- * obtained first in order to prevent deadlock.
- *
- * ip -- the inode being locked
- * lock_flags -- this parameter indicates the inode's locks
- *       to be locked.  It can be:
- *		XFS_IOLOCK_SHARED,
- *		XFS_IOLOCK_EXCL,
- *		XFS_ILOCK_SHARED,
- *		XFS_ILOCK_EXCL,
- *		XFS_IOLOCK_SHARED | XFS_ILOCK_SHARED,
- *		XFS_IOLOCK_SHARED | XFS_ILOCK_EXCL,
- *		XFS_IOLOCK_EXCL | XFS_ILOCK_SHARED,
- *		XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL
- */
-void
-xfs_ilock(
-	xfs_inode_t		*ip,
-	uint			lock_flags)
-{
-	/*
-	 * You can't set both SHARED and EXCL for the same lock,
-	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
-	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
-	 */
-	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
-	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
-	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
-	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
-	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0);
-
-	if (lock_flags & XFS_IOLOCK_EXCL)
-		mrupdate_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags));
-	else if (lock_flags & XFS_IOLOCK_SHARED)
-		mraccess_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags));
-
-	if (lock_flags & XFS_ILOCK_EXCL)
-		mrupdate_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags));
-	else if (lock_flags & XFS_ILOCK_SHARED)
-		mraccess_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags));
-
-	trace_xfs_ilock(ip, lock_flags, _RET_IP_);
-}
-
-/*
- * This is just like xfs_ilock(), except that the caller
- * is guaranteed not to sleep.  It returns 1 if it gets
- * the requested locks and 0 otherwise.  If the IO lock is
- * obtained but the inode lock cannot be, then the IO lock
- * is dropped before returning.
- *
- * ip -- the inode being locked
- * lock_flags -- this parameter indicates the inode's locks to be
- *       to be locked.  See the comment for xfs_ilock() for a list
- *	 of valid values.
- */
-int
-xfs_ilock_nowait(
-	xfs_inode_t		*ip,
-	uint			lock_flags)
-{
-	/*
-	 * You can't set both SHARED and EXCL for the same lock,
-	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
-	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
-	 */
-	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
-	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
-	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
-	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
-	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0);
-
-	if (lock_flags & XFS_IOLOCK_EXCL) {
-		if (!mrtryupdate(&ip->i_iolock))
-			goto out;
-	} else if (lock_flags & XFS_IOLOCK_SHARED) {
-		if (!mrtryaccess(&ip->i_iolock))
-			goto out;
-	}
-	if (lock_flags & XFS_ILOCK_EXCL) {
-		if (!mrtryupdate(&ip->i_lock))
-			goto out_undo_iolock;
-	} else if (lock_flags & XFS_ILOCK_SHARED) {
-		if (!mrtryaccess(&ip->i_lock))
-			goto out_undo_iolock;
-	}
-	trace_xfs_ilock_nowait(ip, lock_flags, _RET_IP_);
-	return 1;
-
- out_undo_iolock:
-	if (lock_flags & XFS_IOLOCK_EXCL)
-		mrunlock_excl(&ip->i_iolock);
-	else if (lock_flags & XFS_IOLOCK_SHARED)
-		mrunlock_shared(&ip->i_iolock);
- out:
-	return 0;
-}
-
-/*
- * xfs_iunlock() is used to drop the inode locks acquired with
- * xfs_ilock() and xfs_ilock_nowait().  The caller must pass
- * in the flags given to xfs_ilock() or xfs_ilock_nowait() so
- * that we know which locks to drop.
- *
- * ip -- the inode being unlocked
- * lock_flags -- this parameter indicates the inode's locks to be
- *       to be unlocked.  See the comment for xfs_ilock() for a list
- *	 of valid values for this parameter.
- *
- */
-void
-xfs_iunlock(
-	xfs_inode_t		*ip,
-	uint			lock_flags)
-{
-	/*
-	 * You can't set both SHARED and EXCL for the same lock,
-	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
-	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
-	 */
-	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
-	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
-	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
-	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
-	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0);
-	ASSERT(lock_flags != 0);
-
-	if (lock_flags & XFS_IOLOCK_EXCL)
-		mrunlock_excl(&ip->i_iolock);
-	else if (lock_flags & XFS_IOLOCK_SHARED)
-		mrunlock_shared(&ip->i_iolock);
-
-	if (lock_flags & XFS_ILOCK_EXCL)
-		mrunlock_excl(&ip->i_lock);
-	else if (lock_flags & XFS_ILOCK_SHARED)
-		mrunlock_shared(&ip->i_lock);
-
-	trace_xfs_iunlock(ip, lock_flags, _RET_IP_);
-}
-
-/*
- * give up write locks.  the i/o lock cannot be held nested
- * if it is being demoted.
- */
-void
-xfs_ilock_demote(
-	xfs_inode_t		*ip,
-	uint			lock_flags)
-{
-	ASSERT(lock_flags & (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL));
-	ASSERT((lock_flags & ~(XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)) == 0);
-
-	if (lock_flags & XFS_ILOCK_EXCL)
-		mrdemote(&ip->i_lock);
-	if (lock_flags & XFS_IOLOCK_EXCL)
-		mrdemote(&ip->i_iolock);
-
-	trace_xfs_ilock_demote(ip, lock_flags, _RET_IP_);
-}
-
-#ifdef DEBUG
-int
-xfs_isilocked(
-	xfs_inode_t		*ip,
-	uint			lock_flags)
-{
-	if (lock_flags & (XFS_ILOCK_EXCL|XFS_ILOCK_SHARED)) {
-		if (!(lock_flags & XFS_ILOCK_SHARED))
-			return !!ip->i_lock.mr_writer;
-		return rwsem_is_locked(&ip->i_lock.mr_lock);
-	}
-
-	if (lock_flags & (XFS_IOLOCK_EXCL|XFS_IOLOCK_SHARED)) {
-		if (!(lock_flags & XFS_IOLOCK_SHARED))
-			return !!ip->i_iolock.mr_writer;
-		return rwsem_is_locked(&ip->i_iolock.mr_lock);
-	}
-
-	ASSERT(0);
-	return 0;
-}
-#endif
-
-void
-__xfs_iflock(
-	struct xfs_inode	*ip)
-{
-	wait_queue_head_t *wq = bit_waitqueue(&ip->i_flags, __XFS_IFLOCK_BIT);
-	DEFINE_WAIT_BIT(wait, &ip->i_flags, __XFS_IFLOCK_BIT);
-
-	do {
-		prepare_to_wait_exclusive(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
-		if (xfs_isiflocked(ip))
-			io_schedule();
-	} while (!xfs_iflock_nowait(ip));
-
-	finish_wait(wq, &wait.wait);
-}

fs/xfs/xfs_inode.c

@@ -45,6 +45,7 @@
 #include "xfs_filestream.h"
 #include "xfs_vnodeops.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 kmem_zone_t *xfs_ifork_zone;
 kmem_zone_t *xfs_inode_zone;
@@ -74,6 +75,256 @@ xfs_get_extsz_hint(
 	return 0;
 }
 
+/*
+ * This is a wrapper routine around the xfs_ilock() routine used to centralize
+ * some grungy code.  It is used in places that wish to lock the inode solely
+ * for reading the extents.  The reason these places can't just call
+ * xfs_ilock(SHARED) is that the inode lock also guards to bringing in of the
+ * extents from disk for a file in b-tree format.  If the inode is in b-tree
+ * format, then we need to lock the inode exclusively until the extents are read
+ * in.  Locking it exclusively all the time would limit our parallelism
+ * unnecessarily, though.  What we do instead is check to see if the extents
+ * have been read in yet, and only lock the inode exclusively if they have not.
+ *
+ * The function returns a value which should be given to the corresponding
+ * xfs_iunlock_map_shared().  This value is the mode in which the lock was
+ * actually taken.
+ */
+uint
+xfs_ilock_map_shared(
+	xfs_inode_t	*ip)
+{
+	uint	lock_mode;
+
+	if ((ip->i_d.di_format == XFS_DINODE_FMT_BTREE) &&
+	    ((ip->i_df.if_flags & XFS_IFEXTENTS) == 0)) {
+		lock_mode = XFS_ILOCK_EXCL;
+	} else {
+		lock_mode = XFS_ILOCK_SHARED;
+	}
+
+	xfs_ilock(ip, lock_mode);
+
+	return lock_mode;
+}
+
+/*
+ * This is simply the unlock routine to go with xfs_ilock_map_shared().
+ * All it does is call xfs_iunlock() with the given lock_mode.
+ */
+void
+xfs_iunlock_map_shared(
+	xfs_inode_t	*ip,
+	unsigned int	lock_mode)
+{
+	xfs_iunlock(ip, lock_mode);
+}
+
+/*
+ * The xfs inode contains 2 locks: a multi-reader lock called the
+ * i_iolock and a multi-reader lock called the i_lock.  This routine
+ * allows either or both of the locks to be obtained.
+ *
+ * The 2 locks should always be ordered so that the IO lock is
+ * obtained first in order to prevent deadlock.
+ *
+ * ip -- the inode being locked
+ * lock_flags -- this parameter indicates the inode's locks
+ *       to be locked.  It can be:
+ *		XFS_IOLOCK_SHARED,
+ *		XFS_IOLOCK_EXCL,
+ *		XFS_ILOCK_SHARED,
+ *		XFS_ILOCK_EXCL,
+ *		XFS_IOLOCK_SHARED | XFS_ILOCK_SHARED,
+ *		XFS_IOLOCK_SHARED | XFS_ILOCK_EXCL,
+ *		XFS_IOLOCK_EXCL | XFS_ILOCK_SHARED,
+ *		XFS_IOLOCK_EXCL | XFS_ILOCK_EXCL
+ */
+void
+xfs_ilock(
+	xfs_inode_t		*ip,
+	uint			lock_flags)
+{
+	trace_xfs_ilock(ip, lock_flags, _RET_IP_);
+
+	/*
+	 * You can't set both SHARED and EXCL for the same lock,
+	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
+	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
+	 */
+	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
+	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
+	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
+	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
+	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0);
+
+	if (lock_flags & XFS_IOLOCK_EXCL)
+		mrupdate_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags));
+	else if (lock_flags & XFS_IOLOCK_SHARED)
+		mraccess_nested(&ip->i_iolock, XFS_IOLOCK_DEP(lock_flags));
+
+	if (lock_flags & XFS_ILOCK_EXCL)
+		mrupdate_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags));
+	else if (lock_flags & XFS_ILOCK_SHARED)
+		mraccess_nested(&ip->i_lock, XFS_ILOCK_DEP(lock_flags));
+}
+
+/*
+ * This is just like xfs_ilock(), except that the caller
+ * is guaranteed not to sleep.  It returns 1 if it gets
+ * the requested locks and 0 otherwise.  If the IO lock is
+ * obtained but the inode lock cannot be, then the IO lock
+ * is dropped before returning.
+ *
+ * ip -- the inode being locked
+ * lock_flags -- this parameter indicates the inode's locks to be
+ *       to be locked.  See the comment for xfs_ilock() for a list
+ *	 of valid values.
+ */
+int
+xfs_ilock_nowait(
+	xfs_inode_t		*ip,
+	uint			lock_flags)
+{
+	trace_xfs_ilock_nowait(ip, lock_flags, _RET_IP_);
+
+	/*
+	 * You can't set both SHARED and EXCL for the same lock,
+	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
+	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
+	 */
+	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
+	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
+	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
+	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
+	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0);
+
+	if (lock_flags & XFS_IOLOCK_EXCL) {
+		if (!mrtryupdate(&ip->i_iolock))
+			goto out;
+	} else if (lock_flags & XFS_IOLOCK_SHARED) {
+		if (!mrtryaccess(&ip->i_iolock))
+			goto out;
+	}
+	if (lock_flags & XFS_ILOCK_EXCL) {
+		if (!mrtryupdate(&ip->i_lock))
+			goto out_undo_iolock;
+	} else if (lock_flags & XFS_ILOCK_SHARED) {
+		if (!mrtryaccess(&ip->i_lock))
+			goto out_undo_iolock;
+	}
+	return 1;
+
+ out_undo_iolock:
+	if (lock_flags & XFS_IOLOCK_EXCL)
+		mrunlock_excl(&ip->i_iolock);
+	else if (lock_flags & XFS_IOLOCK_SHARED)
+		mrunlock_shared(&ip->i_iolock);
+ out:
+	return 0;
+}
+
+/*
+ * xfs_iunlock() is used to drop the inode locks acquired with
+ * xfs_ilock() and xfs_ilock_nowait().  The caller must pass
+ * in the flags given to xfs_ilock() or xfs_ilock_nowait() so
+ * that we know which locks to drop.
+ *
+ * ip -- the inode being unlocked
+ * lock_flags -- this parameter indicates the inode's locks to be
+ *       to be unlocked.  See the comment for xfs_ilock() for a list
+ *	 of valid values for this parameter.
+ *
+ */
+void
+xfs_iunlock(
+	xfs_inode_t		*ip,
+	uint			lock_flags)
+{
+	/*
+	 * You can't set both SHARED and EXCL for the same lock,
+	 * and only XFS_IOLOCK_SHARED, XFS_IOLOCK_EXCL, XFS_ILOCK_SHARED,
+	 * and XFS_ILOCK_EXCL are valid values to set in lock_flags.
+	 */
+	ASSERT((lock_flags & (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL)) !=
+	       (XFS_IOLOCK_SHARED | XFS_IOLOCK_EXCL));
+	ASSERT((lock_flags & (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL)) !=
+	       (XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));
+	ASSERT((lock_flags & ~(XFS_LOCK_MASK | XFS_LOCK_DEP_MASK)) == 0);
+	ASSERT(lock_flags != 0);
+
+	if (lock_flags & XFS_IOLOCK_EXCL)
+		mrunlock_excl(&ip->i_iolock);
+	else if (lock_flags & XFS_IOLOCK_SHARED)
+		mrunlock_shared(&ip->i_iolock);
+
+	if (lock_flags & XFS_ILOCK_EXCL)
+		mrunlock_excl(&ip->i_lock);
+	else if (lock_flags & XFS_ILOCK_SHARED)
+		mrunlock_shared(&ip->i_lock);
+
+	trace_xfs_iunlock(ip, lock_flags, _RET_IP_);
+}
+
+/*
+ * give up write locks.  the i/o lock cannot be held nested
+ * if it is being demoted.
+ */
+void
+xfs_ilock_demote(
+	xfs_inode_t		*ip,
+	uint			lock_flags)
+{
+	ASSERT(lock_flags & (XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL));
+	ASSERT((lock_flags & ~(XFS_IOLOCK_EXCL|XFS_ILOCK_EXCL)) == 0);
+
+	if (lock_flags & XFS_ILOCK_EXCL)
+		mrdemote(&ip->i_lock);
+	if (lock_flags & XFS_IOLOCK_EXCL)
+		mrdemote(&ip->i_iolock);
+
+	trace_xfs_ilock_demote(ip, lock_flags, _RET_IP_);
+}
+
+#ifdef DEBUG
+int
+xfs_isilocked(
+	xfs_inode_t		*ip,
+	uint			lock_flags)
+{
+	if (lock_flags & (XFS_ILOCK_EXCL|XFS_ILOCK_SHARED)) {
+		if (!(lock_flags & XFS_ILOCK_SHARED))
+			return !!ip->i_lock.mr_writer;
+		return rwsem_is_locked(&ip->i_lock.mr_lock);
+	}
+
+	if (lock_flags & (XFS_IOLOCK_EXCL|XFS_IOLOCK_SHARED)) {
+		if (!(lock_flags & XFS_IOLOCK_SHARED))
+			return !!ip->i_iolock.mr_writer;
+		return rwsem_is_locked(&ip->i_iolock.mr_lock);
+	}
+
+	ASSERT(0);
+	return 0;
+}
+#endif
+
+void
+__xfs_iflock(
+	struct xfs_inode	*ip)
+{
+	wait_queue_head_t *wq = bit_waitqueue(&ip->i_flags, __XFS_IFLOCK_BIT);
+	DEFINE_WAIT_BIT(wait, &ip->i_flags, __XFS_IFLOCK_BIT);
+
+	do {
+		prepare_to_wait_exclusive(wq, &wait.wait, TASK_UNINTERRUPTIBLE);
+		if (xfs_isiflocked(ip))
+			io_schedule();
+	} while (!xfs_iflock_nowait(ip));
+
+	finish_wait(wq, &wait.wait);
+}
+
 #ifdef DEBUG
 /*
  * Make sure that the extents in the given memory buffer
@@ -131,6 +382,65 @@ xfs_inobp_check(
 }
 #endif
 
+static void
+xfs_inode_buf_verify(
+	struct xfs_buf	*bp)
+{
+	struct xfs_mount *mp = bp->b_target->bt_mount;
+	int		i;
+	int		ni;
+
+	/*
+	 * Validate the magic number and version of every inode in the buffer
+	 */
+	ni = XFS_BB_TO_FSB(mp, bp->b_length) * mp->m_sb.sb_inopblock;
+	for (i = 0; i < ni; i++) {
+		int		di_ok;
+		xfs_dinode_t	*dip;
+
+		dip = (struct xfs_dinode *)xfs_buf_offset(bp,
+					(i << mp->m_sb.sb_inodelog));
+		di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
+			    XFS_DINODE_GOOD_VERSION(dip->di_version);
+		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
+						XFS_ERRTAG_ITOBP_INOTOBP,
+						XFS_RANDOM_ITOBP_INOTOBP))) {
+			xfs_buf_ioerror(bp, EFSCORRUPTED);
+			XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_HIGH,
+					     mp, dip);
+#ifdef DEBUG
+			xfs_emerg(mp,
+				"bad inode magic/vsn daddr %lld #%d (magic=%x)",
+				(unsigned long long)bp->b_bn, i,
+				be16_to_cpu(dip->di_magic));
+			ASSERT(0);
+#endif
+		}
+	}
+	xfs_inobp_check(mp, bp);
+}
+
+
+static void
+xfs_inode_buf_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_inode_buf_verify(bp);
+}
+
+static void
+xfs_inode_buf_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_inode_buf_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_inode_buf_ops = {
+	.verify_read = xfs_inode_buf_read_verify,
+	.verify_write = xfs_inode_buf_write_verify,
+};
+
+
 /*
  * This routine is called to map an inode to the buffer containing the on-disk
  * version of the inode.  It returns a pointer to the buffer containing the
@@ -145,71 +455,33 @@ xfs_imap_to_bp(
 	struct xfs_mount	*mp,
 	struct xfs_trans	*tp,
 	struct xfs_imap		*imap,
-	struct xfs_dinode	**dipp,
+	struct xfs_dinode       **dipp,
 	struct xfs_buf		**bpp,
 	uint			buf_flags,
 	uint			iget_flags)
 {
 	struct xfs_buf		*bp;
 	int			error;
-	int			i;
-	int			ni;
 
 	buf_flags |= XBF_UNMAPPED;
 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp, imap->im_blkno,
-				   (int)imap->im_len, buf_flags, &bp);
+				   (int)imap->im_len, buf_flags, &bp,
+				   &xfs_inode_buf_ops);
 	if (error) {
-		if (error != EAGAIN) {
-			xfs_warn(mp,
-				"%s: xfs_trans_read_buf() returned error %d.",
-				__func__, error);
-		} else {
+		if (error == EAGAIN) {
 			ASSERT(buf_flags & XBF_TRYLOCK);
+			return error;
 		}
-		return error;
-	}
-
-	/*
-	 * Validate the magic number and version of every inode in the buffer
-	 * (if DEBUG kernel) or the first inode in the buffer, otherwise.
-	 */
-#ifdef DEBUG
-	ni = BBTOB(imap->im_len) >> mp->m_sb.sb_inodelog;
-#else	/* usual case */
-	ni = 1;
-#endif
 
-	for (i = 0; i < ni; i++) {
-		int		di_ok;
-		xfs_dinode_t	*dip;
+		if (error == EFSCORRUPTED &&
+		    (iget_flags & XFS_IGET_UNTRUSTED))
+			return XFS_ERROR(EINVAL);
 
-		dip = (xfs_dinode_t *)xfs_buf_offset(bp,
-					(i << mp->m_sb.sb_inodelog));
-		di_ok = dip->di_magic == cpu_to_be16(XFS_DINODE_MAGIC) &&
-			    XFS_DINODE_GOOD_VERSION(dip->di_version);
-		if (unlikely(XFS_TEST_ERROR(!di_ok, mp,
-						XFS_ERRTAG_ITOBP_INOTOBP,
-						XFS_RANDOM_ITOBP_INOTOBP))) {
-			if (iget_flags & XFS_IGET_UNTRUSTED) {
-				xfs_trans_brelse(tp, bp);
-				return XFS_ERROR(EINVAL);
-			}
-			XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_HIGH,
-					     mp, dip);
-#ifdef DEBUG
-			xfs_emerg(mp,
-				"bad inode magic/vsn daddr %lld #%d (magic=%x)",
-				(unsigned long long)imap->im_blkno, i,
-				be16_to_cpu(dip->di_magic));
-			ASSERT(0);
-#endif
-			xfs_trans_brelse(tp, bp);
-			return XFS_ERROR(EFSCORRUPTED);
-		}
+		xfs_warn(mp, "%s: xfs_trans_read_buf() returned error %d.",
+			__func__, error);
+		return error;
 	}
 
-	xfs_inobp_check(mp, bp);
-
 	*bpp = bp;
 	*dipp = (struct xfs_dinode *)xfs_buf_offset(bp, imap->im_boffset);
 	return 0;
@@ -853,16 +1125,16 @@ xfs_iread_extents(
  * set according to the contents of the given cred structure.
  *
  * Use xfs_dialloc() to allocate the on-disk inode. If xfs_dialloc()
- * has a free inode available, call xfs_iget()
- * to obtain the in-core version of the allocated inode.  Finally,
- * fill in the inode and log its initial contents.  In this case,
- * ialloc_context would be set to NULL and call_again set to false.
+ * has a free inode available, call xfs_iget() to obtain the in-core
+ * version of the allocated inode.  Finally, fill in the inode and
+ * log its initial contents.  In this case, ialloc_context would be
+ * set to NULL.
  *
- * If xfs_dialloc() does not have an available inode,
- * it will replenish its supply by doing an allocation. Since we can
- * only do one allocation within a transaction without deadlocks, we
- * must commit the current transaction before returning the inode itself.
- * In this case, therefore, we will set call_again to true and return.
+ * If xfs_dialloc() does not have an available inode, it will replenish
+ * its supply by doing an allocation. Since we can only do one
+ * allocation within a transaction without deadlocks, we must commit
+ * the current transaction before returning the inode itself.
+ * In this case, therefore, we will set ialloc_context and return.
  * The caller should then commit the current transaction, start a new
  * transaction, and call xfs_ialloc() again to actually get the inode.
  *
@@ -1514,6 +1786,18 @@ xfs_ifree_cluster(
 
 		if (!bp)
 			return ENOMEM;
+
+		/*
+		 * This buffer may not have been correctly initialised as we
+		 * didn't read it from disk. That's not important because we are
+		 * only using to mark the buffer as stale in the log, and to
+		 * attach stale cached inodes on it. That means it will never be
+		 * dispatched for IO. If it is, we want to know about it, and we
+		 * want it to fail. We can acheive this by adding a write
+		 * verifier to the buffer.
+		 */
+		 bp->b_ops = &xfs_inode_buf_ops;
+
 		/*
 		 * Walk the inodes already attached to the buffer and mark them
 		 * stale. These will all have the flush locks held, so an
@@ -3661,3 +3945,40 @@ xfs_iext_irec_update_extoffs(
 		ifp->if_u1.if_ext_irec[i].er_extoff += ext_diff;
 	}
 }
+
+/*
+ * Test whether it is appropriate to check an inode for and free post EOF
+ * blocks. The 'force' parameter determines whether we should also consider
+ * regular files that are marked preallocated or append-only.
+ */
+bool
+xfs_can_free_eofblocks(struct xfs_inode *ip, bool force)
+{
+	/* prealloc/delalloc exists only on regular files */
+	if (!S_ISREG(ip->i_d.di_mode))
+		return false;
+
+	/*
+	 * Zero sized files with no cached pages and delalloc blocks will not
+	 * have speculative prealloc/delalloc blocks to remove.
+	 */
+	if (VFS_I(ip)->i_size == 0 &&
+	    VN_CACHED(VFS_I(ip)) == 0 &&
+	    ip->i_delayed_blks == 0)
+		return false;
+
+	/* If we haven't read in the extent list, then don't do it now. */
+	if (!(ip->i_df.if_flags & XFS_IFEXTENTS))
+		return false;
+
+	/*
+	 * Do not free real preallocated or append-only files unless the file
+	 * has delalloc blocks and we are forced to remove them.
+	 */
+	if (ip->i_d.di_flags & (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND))
+		if (!force || ip->i_delayed_blks == 0)
+			return false;
+
+	return true;
+}
+

fs/xfs/xfs_inode.h

@@ -496,11 +496,10 @@ static inline int xfs_isiflocked(struct xfs_inode *ip)
 	(((pip)->i_mount->m_flags & XFS_MOUNT_GRPID) || \
 	 ((pip)->i_d.di_mode & S_ISGID))
 
+
 /*
- * xfs_iget.c prototypes.
+ * xfs_inode.c prototypes.
  */
-int		xfs_iget(struct xfs_mount *, struct xfs_trans *, xfs_ino_t,
-			 uint, uint, xfs_inode_t **);
 void		xfs_ilock(xfs_inode_t *, uint);
 int		xfs_ilock_nowait(xfs_inode_t *, uint);
 void		xfs_iunlock(xfs_inode_t *, uint);
@@ -508,11 +507,6 @@ void		xfs_ilock_demote(xfs_inode_t *, uint);
 int		xfs_isilocked(xfs_inode_t *, uint);
 uint		xfs_ilock_map_shared(xfs_inode_t *);
 void		xfs_iunlock_map_shared(xfs_inode_t *, uint);
-void		xfs_inode_free(struct xfs_inode *ip);
-
-/*
- * xfs_inode.c prototypes.
- */
 int		xfs_ialloc(struct xfs_trans *, xfs_inode_t *, umode_t,
 			   xfs_nlink_t, xfs_dev_t, prid_t, int,
 			   struct xfs_buf **, xfs_inode_t **);
@@ -591,6 +585,7 @@ void		xfs_iext_irec_compact(xfs_ifork_t *);
 void		xfs_iext_irec_compact_pages(xfs_ifork_t *);
 void		xfs_iext_irec_compact_full(xfs_ifork_t *);
 void		xfs_iext_irec_update_extoffs(xfs_ifork_t *, int, int);
+bool		xfs_can_free_eofblocks(struct xfs_inode *, bool);
 
 #define xfs_ipincount(ip)	((unsigned int) atomic_read(&ip->i_pincount))
 
@@ -603,5 +598,6 @@ void		xfs_inobp_check(struct xfs_mount *, struct xfs_buf *);
 extern struct kmem_zone	*xfs_ifork_zone;
 extern struct kmem_zone	*xfs_inode_zone;
 extern struct kmem_zone	*xfs_ili_zone;
+extern const struct xfs_buf_ops xfs_inode_buf_ops;
 
 #endif	/* __XFS_INODE_H__ */

fs/xfs/xfs_ioctl.c

@@ -42,6 +42,7 @@
 #include "xfs_inode_item.h"
 #include "xfs_export.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 #include <linux/capability.h>
 #include <linux/dcache.h>
@@ -1602,6 +1603,26 @@ xfs_file_ioctl(
 		error = xfs_errortag_clearall(mp, 1);
 		return -error;
 
+	case XFS_IOC_FREE_EOFBLOCKS: {
+		struct xfs_eofblocks eofb;
+
+		if (copy_from_user(&eofb, arg, sizeof(eofb)))
+			return -XFS_ERROR(EFAULT);
+
+		if (eofb.eof_version != XFS_EOFBLOCKS_VERSION)
+			return -XFS_ERROR(EINVAL);
+
+		if (eofb.eof_flags & ~XFS_EOF_FLAGS_VALID)
+			return -XFS_ERROR(EINVAL);
+
+		if (memchr_inv(&eofb.pad32, 0, sizeof(eofb.pad32)) ||
+		    memchr_inv(eofb.pad64, 0, sizeof(eofb.pad64)))
+			return -XFS_ERROR(EINVAL);
+
+		error = xfs_icache_free_eofblocks(mp, &eofb);
+		return -error;
+	}
+
 	default:
 		return -ENOTTY;
 	}

fs/xfs/xfs_iomap.c

@@ -41,6 +41,7 @@
 #include "xfs_utils.h"
 #include "xfs_iomap.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 
 #define XFS_WRITEIO_ALIGN(mp,off)	(((off) >> mp->m_writeio_log) \
@@ -373,7 +374,7 @@ xfs_iomap_write_delay(
 	xfs_extlen_t	extsz;
 	int		nimaps;
 	xfs_bmbt_irec_t imap[XFS_WRITE_IMAPS];
-	int		prealloc, flushed = 0;
+	int		prealloc;
 	int		error;
 
 	ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
@@ -434,31 +435,29 @@ retry:
 	}
 
 	/*
-	 * If bmapi returned us nothing, we got either ENOSPC or EDQUOT.  For
-	 * ENOSPC, * flush all other inodes with delalloc blocks to free up
-	 * some of the excess reserved metadata space. For both cases, retry
+	 * If bmapi returned us nothing, we got either ENOSPC or EDQUOT. Retry
 	 * without EOF preallocation.
 	 */
 	if (nimaps == 0) {
 		trace_xfs_delalloc_enospc(ip, offset, count);
-		if (flushed)
-			return XFS_ERROR(error ? error : ENOSPC);
-
-		if (error == ENOSPC) {
-			xfs_iunlock(ip, XFS_ILOCK_EXCL);
-			xfs_flush_inodes(ip);
-			xfs_ilock(ip, XFS_ILOCK_EXCL);
+		if (prealloc) {
+			prealloc = 0;
+			error = 0;
+			goto retry;
 		}
-
-		flushed = 1;
-		error = 0;
-		prealloc = 0;
-		goto retry;
+		return XFS_ERROR(error ? error : ENOSPC);
 	}
 
 	if (!(imap[0].br_startblock || XFS_IS_REALTIME_INODE(ip)))
 		return xfs_alert_fsblock_zero(ip, &imap[0]);
 
+	/*
+	 * Tag the inode as speculatively preallocated so we can reclaim this
+	 * space on demand, if necessary.
+	 */
+	if (prealloc)
+		xfs_inode_set_eofblocks_tag(ip);
+
 	*ret_imap = imap[0];
 	return 0;
 }

fs/xfs/xfs_iops.c

@@ -38,6 +38,7 @@
 #include "xfs_vnodeops.h"
 #include "xfs_inode_item.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 #include <linux/capability.h>
 #include <linux/xattr.h>
@@ -779,8 +780,8 @@ xfs_setattr_size(
 	 * care about here.
 	 */
 	if (oldsize != ip->i_d.di_size && newsize > ip->i_d.di_size) {
-		error = xfs_flush_pages(ip, ip->i_d.di_size, newsize, 0,
-					FI_NONE);
+		error = -filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
+						      ip->i_d.di_size, newsize);
 		if (error)
 			goto out_unlock;
 	}
@@ -854,6 +855,9 @@ xfs_setattr_size(
 		 * and do not wait the usual (long) time for writeout.
 		 */
 		xfs_iflags_set(ip, XFS_ITRUNCATED);
+
+		/* A truncate down always removes post-EOF blocks. */
+		xfs_inode_clear_eofblocks_tag(ip);
 	}
 
 	if (mask & ATTR_CTIME) {

fs/xfs/xfs_itable.c

@@ -34,6 +34,7 @@
 #include "xfs_error.h"
 #include "xfs_btree.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 STATIC int
 xfs_internal_inum(
@@ -395,7 +396,8 @@ xfs_bulkstat(
 					if (xfs_inobt_maskn(chunkidx, nicluster)
 							& ~r.ir_free)
 						xfs_btree_reada_bufs(mp, agno,
-							agbno, nbcluster);
+							agbno, nbcluster,
+							&xfs_inode_buf_ops);
 				}
 				irbp->ir_startino = r.ir_startino;
 				irbp->ir_freecount = r.ir_freecount;

fs/xfs/xfs_linux.h

@@ -44,6 +44,7 @@
 #include <linux/kernel.h>
 #include <linux/blkdev.h>
 #include <linux/slab.h>
+#include <linux/crc32c.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
 #include <linux/file.h>
@@ -118,6 +119,7 @@
 #define xfs_rotorstep		xfs_params.rotorstep.val
 #define xfs_inherit_nodefrag	xfs_params.inherit_nodfrg.val
 #define xfs_fstrm_centisecs	xfs_params.fstrm_timer.val
+#define xfs_eofb_secs		xfs_params.eofb_timer.val
 
 #define current_cpu()		(raw_smp_processor_id())
 #define current_pid()		(current->pid)

fs/xfs/xfs_log.c

@@ -34,6 +34,8 @@
 #include "xfs_dinode.h"
 #include "xfs_inode.h"
 #include "xfs_trace.h"
+#include "xfs_fsops.h"
+#include "xfs_cksum.h"
 
 kmem_zone_t	*xfs_log_ticket_zone;
 
@@ -458,7 +460,8 @@ xfs_log_reserve(
 	tic->t_trans_type = t_type;
 	*ticp = tic;
 
-	xlog_grant_push_ail(log, tic->t_unit_res * tic->t_cnt);
+	xlog_grant_push_ail(log, tic->t_cnt ? tic->t_unit_res * tic->t_cnt
+					    : tic->t_unit_res);
 
 	trace_xfs_log_reserve(log, tic);
 
@@ -679,25 +682,29 @@ out:
 }
 
 /*
- * Finish the recovery of the file system.  This is separate from
- * the xfs_log_mount() call, because it depends on the code in
- * xfs_mountfs() to read in the root and real-time bitmap inodes
- * between calling xfs_log_mount() and here.
+ * Finish the recovery of the file system.  This is separate from the
+ * xfs_log_mount() call, because it depends on the code in xfs_mountfs() to read
+ * in the root and real-time bitmap inodes between calling xfs_log_mount() and
+ * here.
  *
- * mp		- ubiquitous xfs mount point structure
+ * If we finish recovery successfully, start the background log work. If we are
+ * not doing recovery, then we have a RO filesystem and we don't need to start
+ * it.
  */
 int
 xfs_log_mount_finish(xfs_mount_t *mp)
 {
-	int	error;
+	int	error = 0;
 
-	if (!(mp->m_flags & XFS_MOUNT_NORECOVERY))
+	if (!(mp->m_flags & XFS_MOUNT_NORECOVERY)) {
 		error = xlog_recover_finish(mp->m_log);
-	else {
-		error = 0;
+		if (!error)
+			xfs_log_work_queue(mp);
+	} else {
 		ASSERT(mp->m_flags & XFS_MOUNT_RDONLY);
 	}
 
+
 	return error;
 }
 
@@ -850,15 +857,49 @@ xfs_log_unmount_write(xfs_mount_t *mp)
 }	/* xfs_log_unmount_write */
 
 /*
- * Deallocate log structures for unmount/relocation.
+ * Empty the log for unmount/freeze.
+ *
+ * To do this, we first need to shut down the background log work so it is not
+ * trying to cover the log as we clean up. We then need to unpin all objects in
+ * the log so we can then flush them out. Once they have completed their IO and
+ * run the callbacks removing themselves from the AIL, we can write the unmount
+ * record.
+ */
+void
+xfs_log_quiesce(
+	struct xfs_mount	*mp)
+{
+	cancel_delayed_work_sync(&mp->m_log->l_work);
+	xfs_log_force(mp, XFS_LOG_SYNC);
+
+	/*
+	 * The superblock buffer is uncached and while xfs_ail_push_all_sync()
+	 * will push it, xfs_wait_buftarg() will not wait for it. Further,
+	 * xfs_buf_iowait() cannot be used because it was pushed with the
+	 * XBF_ASYNC flag set, so we need to use a lock/unlock pair to wait for
+	 * the IO to complete.
+	 */
+	xfs_ail_push_all_sync(mp->m_ail);
+	xfs_wait_buftarg(mp->m_ddev_targp);
+	xfs_buf_lock(mp->m_sb_bp);
+	xfs_buf_unlock(mp->m_sb_bp);
+
+	xfs_log_unmount_write(mp);
+}
+
+/*
+ * Shut down and release the AIL and Log.
  *
- * We need to stop the aild from running before we destroy
- * and deallocate the log as the aild references the log.
+ * During unmount, we need to ensure we flush all the dirty metadata objects
+ * from the AIL so that the log is empty before we write the unmount record to
+ * the log. Once this is done, we can tear down the AIL and the log.
  */
 void
-xfs_log_unmount(xfs_mount_t *mp)
+xfs_log_unmount(
+	struct xfs_mount	*mp)
 {
-	cancel_delayed_work_sync(&mp->m_sync_work);
+	xfs_log_quiesce(mp);
+
 	xfs_trans_ail_destroy(mp);
 	xlog_dealloc_log(mp->m_log);
 }
@@ -1090,8 +1131,7 @@ xlog_iodone(xfs_buf_t *bp)
 	 * with it being freed after writing the unmount record to the
 	 * log.
 	 */
-
-}	/* xlog_iodone */
+}
 
 /*
  * Return size of each in-core log record buffer.
@@ -1161,6 +1201,40 @@ done:
 }	/* xlog_get_iclog_buffer_size */
 
 
+void
+xfs_log_work_queue(
+	struct xfs_mount        *mp)
+{
+	queue_delayed_work(mp->m_log_workqueue, &mp->m_log->l_work,
+				msecs_to_jiffies(xfs_syncd_centisecs * 10));
+}
+
+/*
+ * Every sync period we need to unpin all items in the AIL and push them to
+ * disk. If there is nothing dirty, then we might need to cover the log to
+ * indicate that the filesystem is idle.
+ */
+void
+xfs_log_worker(
+	struct work_struct	*work)
+{
+	struct xlog		*log = container_of(to_delayed_work(work),
+						struct xlog, l_work);
+	struct xfs_mount	*mp = log->l_mp;
+
+	/* dgc: errors ignored - not fatal and nowhere to report them */
+	if (xfs_log_need_covered(mp))
+		xfs_fs_log_dummy(mp);
+	else
+		xfs_log_force(mp, 0);
+
+	/* start pushing all the metadata that is currently dirty */
+	xfs_ail_push_all(mp->m_ail);
+
+	/* queue us up again */
+	xfs_log_work_queue(mp);
+}
+
 /*
  * This routine initializes some of the log structure for a given mount point.
  * Its primary purpose is to fill in enough, so recovery can occur.  However,
@@ -1195,6 +1269,7 @@ xlog_alloc_log(
 	log->l_logBBsize   = num_bblks;
 	log->l_covered_state = XLOG_STATE_COVER_IDLE;
 	log->l_flags	   |= XLOG_ACTIVE_RECOVERY;
+	INIT_DELAYED_WORK(&log->l_work, xfs_log_worker);
 
 	log->l_prev_block  = -1;
 	/* log->l_tail_lsn = 0x100000000LL; cycle = 1; current block = 0 */
@@ -1416,6 +1491,84 @@ xlog_grant_push_ail(
 		xfs_ail_push(log->l_ailp, threshold_lsn);
 }
 
+/*
+ * Stamp cycle number in every block
+ */
+STATIC void
+xlog_pack_data(
+	struct xlog		*log,
+	struct xlog_in_core	*iclog,
+	int			roundoff)
+{
+	int			i, j, k;
+	int			size = iclog->ic_offset + roundoff;
+	__be32			cycle_lsn;
+	xfs_caddr_t		dp;
+
+	cycle_lsn = CYCLE_LSN_DISK(iclog->ic_header.h_lsn);
+
+	dp = iclog->ic_datap;
+	for (i = 0; i < BTOBB(size); i++) {
+		if (i >= (XLOG_HEADER_CYCLE_SIZE / BBSIZE))
+			break;
+		iclog->ic_header.h_cycle_data[i] = *(__be32 *)dp;
+		*(__be32 *)dp = cycle_lsn;
+		dp += BBSIZE;
+	}
+
+	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
+		xlog_in_core_2_t *xhdr = iclog->ic_data;
+
+		for ( ; i < BTOBB(size); i++) {
+			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
+			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
+			xhdr[j].hic_xheader.xh_cycle_data[k] = *(__be32 *)dp;
+			*(__be32 *)dp = cycle_lsn;
+			dp += BBSIZE;
+		}
+
+		for (i = 1; i < log->l_iclog_heads; i++)
+			xhdr[i].hic_xheader.xh_cycle = cycle_lsn;
+	}
+}
+
+/*
+ * Calculate the checksum for a log buffer.
+ *
+ * This is a little more complicated than it should be because the various
+ * headers and the actual data are non-contiguous.
+ */
+__le32
+xlog_cksum(
+	struct xlog		*log,
+	struct xlog_rec_header	*rhead,
+	char			*dp,
+	int			size)
+{
+	__uint32_t		crc;
+
+	/* first generate the crc for the record header ... */
+	crc = xfs_start_cksum((char *)rhead,
+			      sizeof(struct xlog_rec_header),
+			      offsetof(struct xlog_rec_header, h_crc));
+
+	/* ... then for additional cycle data for v2 logs ... */
+	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
+		union xlog_in_core2 *xhdr = (union xlog_in_core2 *)rhead;
+		int		i;
+
+		for (i = 1; i < log->l_iclog_heads; i++) {
+			crc = crc32c(crc, &xhdr[i].hic_xheader,
+				     sizeof(struct xlog_rec_ext_header));
+		}
+	}
+
+	/* ... and finally for the payload */
+	crc = crc32c(crc, dp, size);
+
+	return xfs_end_cksum(crc);
+}
+
 /*
  * The bdstrat callback function for log bufs. This gives us a central
  * place to trap bufs in case we get hit by a log I/O error and need to
@@ -1476,7 +1629,6 @@ xlog_sync(
 	struct xlog		*log,
 	struct xlog_in_core	*iclog)
 {
-	xfs_caddr_t	dptr;		/* pointer to byte sized element */
 	xfs_buf_t	*bp;
 	int		i;
 	uint		count;		/* byte count of bwrite */
@@ -1485,6 +1637,7 @@ xlog_sync(
 	int		split = 0;	/* split write into two regions */
 	int		error;
 	int		v2 = xfs_sb_version_haslogv2(&log->l_mp->m_sb);
+	int		size;
 
 	XFS_STATS_INC(xs_log_writes);
 	ASSERT(atomic_read(&iclog->ic_refcnt) == 0);
@@ -1515,13 +1668,10 @@ xlog_sync(
 	xlog_pack_data(log, iclog, roundoff); 
 
 	/* real byte length */
-	if (v2) {
-		iclog->ic_header.h_len =
-			cpu_to_be32(iclog->ic_offset + roundoff);
-	} else {
-		iclog->ic_header.h_len =
-			cpu_to_be32(iclog->ic_offset);
-	}
+	size = iclog->ic_offset;
+	if (v2)
+		size += roundoff;
+	iclog->ic_header.h_len = cpu_to_be32(size);
 
 	bp = iclog->ic_bp;
 	XFS_BUF_SET_ADDR(bp, BLOCK_LSN(be64_to_cpu(iclog->ic_header.h_lsn)));
@@ -1530,12 +1680,36 @@ xlog_sync(
 
 	/* Do we need to split this write into 2 parts? */
 	if (XFS_BUF_ADDR(bp) + BTOBB(count) > log->l_logBBsize) {
+		char		*dptr;
+
 		split = count - (BBTOB(log->l_logBBsize - XFS_BUF_ADDR(bp)));
 		count = BBTOB(log->l_logBBsize - XFS_BUF_ADDR(bp));
-		iclog->ic_bwritecnt = 2;	/* split into 2 writes */
+		iclog->ic_bwritecnt = 2;
+
+		/*
+		 * Bump the cycle numbers at the start of each block in the
+		 * part of the iclog that ends up in the buffer that gets
+		 * written to the start of the log.
+		 *
+		 * Watch out for the header magic number case, though.
+		 */
+		dptr = (char *)&iclog->ic_header + count;
+		for (i = 0; i < split; i += BBSIZE) {
+			__uint32_t cycle = be32_to_cpu(*(__be32 *)dptr);
+			if (++cycle == XLOG_HEADER_MAGIC_NUM)
+				cycle++;
+			*(__be32 *)dptr = cpu_to_be32(cycle);
+
+			dptr += BBSIZE;
+		}
 	} else {
 		iclog->ic_bwritecnt = 1;
 	}
+
+	/* calculcate the checksum */
+	iclog->ic_header.h_crc = xlog_cksum(log, &iclog->ic_header,
+					    iclog->ic_datap, size);
+
 	bp->b_io_length = BTOBB(count);
 	bp->b_fspriv = iclog;
 	XFS_BUF_ZEROFLAGS(bp);
@@ -1589,19 +1763,6 @@ xlog_sync(
 		bp->b_flags |= XBF_SYNCIO;
 		if (log->l_mp->m_flags & XFS_MOUNT_BARRIER)
 			bp->b_flags |= XBF_FUA;
-		dptr = bp->b_addr;
-		/*
-		 * Bump the cycle numbers at the start of each block
-		 * since this part of the buffer is at the start of
-		 * a new cycle.  Watch out for the header magic number
-		 * case, though.
-		 */
-		for (i = 0; i < split; i += BBSIZE) {
-			be32_add_cpu((__be32 *)dptr, 1);
-			if (be32_to_cpu(*(__be32 *)dptr) == XLOG_HEADER_MAGIC_NUM)
-				be32_add_cpu((__be32 *)dptr, 1);
-			dptr += BBSIZE;
-		}
 
 		ASSERT(XFS_BUF_ADDR(bp) <= log->l_logBBsize-1);
 		ASSERT(XFS_BUF_ADDR(bp) + BTOBB(count) <= log->l_logBBsize);
@@ -1618,7 +1779,6 @@ xlog_sync(
 	return 0;
 }	/* xlog_sync */
 
-
 /*
  * Deallocate a log structure
  */
@@ -3713,3 +3873,4 @@ xlog_iclogs_empty(
 	} while (iclog != log->l_iclog);
 	return 1;
 }
+

fs/xfs/xfs_log.h

@@ -181,5 +181,9 @@ int	xfs_log_commit_cil(struct xfs_mount *mp, struct xfs_trans *tp,
 				xfs_lsn_t *commit_lsn, int flags);
 bool	xfs_log_item_in_current_chkpt(struct xfs_log_item *lip);
 
+void	xfs_log_work_queue(struct xfs_mount *mp);
+void	xfs_log_worker(struct work_struct *work);
+void	xfs_log_quiesce(struct xfs_mount *mp);
+
 #endif
 #endif	/* __XFS_LOG_H__ */

fs/xfs/xfs_log_priv.h

@@ -139,7 +139,6 @@ static inline uint xlog_get_client_id(__be32 i)
 /*
  * Flags for log structure
  */
-#define XLOG_CHKSUM_MISMATCH	0x1	/* used only during recovery */
 #define XLOG_ACTIVE_RECOVERY	0x2	/* in the middle of recovery */
 #define	XLOG_RECOVERY_NEEDED	0x4	/* log was recovered */
 #define XLOG_IO_ERROR		0x8	/* log hit an I/O error, and being
@@ -291,7 +290,7 @@ typedef struct xlog_rec_header {
 	__be32	  h_len;	/* len in bytes; should be 64-bit aligned: 4 */
 	__be64	  h_lsn;	/* lsn of this LR			:  8 */
 	__be64	  h_tail_lsn;	/* lsn of 1st LR w/ buffers not committed: 8 */
-	__be32	  h_chksum;	/* may not be used; non-zero if used	:  4 */
+	__le32	  h_crc;	/* crc of log record                    :  4 */
 	__be32	  h_prev_block; /* block number to previous LR		:  4 */
 	__be32	  h_num_logops;	/* number of log operations in this LR	:  4 */
 	__be32	  h_cycle_data[XLOG_HEADER_CYCLE_SIZE / BBSIZE];
@@ -495,6 +494,7 @@ struct xlog {
 	struct xfs_buf		*l_xbuf;        /* extra buffer for log
 						 * wrapping */
 	struct xfs_buftarg	*l_targ;        /* buftarg of log */
+	struct delayed_work	l_work;		/* background flush work */
 	uint			l_flags;
 	uint			l_quotaoffs_flag; /* XFS_DQ_*, for QUOTAOFFs */
 	struct list_head	*l_buf_cancel_table;
@@ -554,11 +554,9 @@ xlog_recover(
 extern int
 xlog_recover_finish(
 	struct xlog		*log);
-extern void
-xlog_pack_data(
-	struct xlog		*log,
-	struct xlog_in_core	*iclog,
-	int);
+
+extern __le32	 xlog_cksum(struct xlog *log, struct xlog_rec_header *rhead,
+			    char *dp, int size);
 
 extern kmem_zone_t *xfs_log_ticket_zone;
 struct xlog_ticket *

fs/xfs/xfs_log_recover.c

@@ -41,7 +41,9 @@
 #include "xfs_trans_priv.h"
 #include "xfs_quota.h"
 #include "xfs_utils.h"
+#include "xfs_cksum.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 STATIC int
 xlog_find_zeroed(
@@ -2143,7 +2145,7 @@ xlog_recover_buffer_pass2(
 		buf_flags |= XBF_UNMAPPED;
 
 	bp = xfs_buf_read(mp->m_ddev_targp, buf_f->blf_blkno, buf_f->blf_len,
-			  buf_flags);
+			  buf_flags, NULL);
 	if (!bp)
 		return XFS_ERROR(ENOMEM);
 	error = bp->b_error;
@@ -2236,7 +2238,8 @@ xlog_recover_inode_pass2(
 	}
 	trace_xfs_log_recover_inode_recover(log, in_f);
 
-	bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0);
+	bp = xfs_buf_read(mp->m_ddev_targp, in_f->ilf_blkno, in_f->ilf_len, 0,
+			  NULL);
 	if (!bp) {
 		error = ENOMEM;
 		goto error;
@@ -2547,7 +2550,8 @@ xlog_recover_dquot_pass2(
 	ASSERT(dq_f->qlf_len == 1);
 
 	error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp, dq_f->qlf_blkno,
-				   XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp);
+				   XFS_FSB_TO_BB(mp, dq_f->qlf_len), 0, &bp,
+				   NULL);
 	if (error)
 		return error;
 
@@ -3213,80 +3217,58 @@ xlog_recover_process_iunlinks(
 	mp->m_dmevmask = mp_dmevmask;
 }
 
-
-#ifdef DEBUG
-STATIC void
-xlog_pack_data_checksum(
-	struct xlog		*log,
-	struct xlog_in_core	*iclog,
-	int			size)
-{
-	int		i;
-	__be32		*up;
-	uint		chksum = 0;
-
-	up = (__be32 *)iclog->ic_datap;
-	/* divide length by 4 to get # words */
-	for (i = 0; i < (size >> 2); i++) {
-		chksum ^= be32_to_cpu(*up);
-		up++;
-	}
-	iclog->ic_header.h_chksum = cpu_to_be32(chksum);
-}
-#else
-#define xlog_pack_data_checksum(log, iclog, size)
-#endif
-
 /*
- * Stamp cycle number in every block
+ * Upack the log buffer data and crc check it. If the check fails, issue a
+ * warning if and only if the CRC in the header is non-zero. This makes the
+ * check an advisory warning, and the zero CRC check will prevent failure
+ * warnings from being emitted when upgrading the kernel from one that does not
+ * add CRCs by default.
+ *
+ * When filesystems are CRC enabled, this CRC mismatch becomes a fatal log
+ * corruption failure
  */
-void
-xlog_pack_data(
-	struct xlog		*log,
-	struct xlog_in_core	*iclog,
-	int			roundoff)
+STATIC int
+xlog_unpack_data_crc(
+	struct xlog_rec_header	*rhead,
+	xfs_caddr_t		dp,
+	struct xlog		*log)
 {
-	int			i, j, k;
-	int			size = iclog->ic_offset + roundoff;
-	__be32			cycle_lsn;
-	xfs_caddr_t		dp;
-
-	xlog_pack_data_checksum(log, iclog, size);
-
-	cycle_lsn = CYCLE_LSN_DISK(iclog->ic_header.h_lsn);
-
-	dp = iclog->ic_datap;
-	for (i = 0; i < BTOBB(size) &&
-		i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
-		iclog->ic_header.h_cycle_data[i] = *(__be32 *)dp;
-		*(__be32 *)dp = cycle_lsn;
-		dp += BBSIZE;
-	}
-
-	if (xfs_sb_version_haslogv2(&log->l_mp->m_sb)) {
-		xlog_in_core_2_t *xhdr = iclog->ic_data;
-
-		for ( ; i < BTOBB(size); i++) {
-			j = i / (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
-			k = i % (XLOG_HEADER_CYCLE_SIZE / BBSIZE);
-			xhdr[j].hic_xheader.xh_cycle_data[k] = *(__be32 *)dp;
-			*(__be32 *)dp = cycle_lsn;
-			dp += BBSIZE;
+	__le32			crc;
+
+	crc = xlog_cksum(log, rhead, dp, be32_to_cpu(rhead->h_len));
+	if (crc != rhead->h_crc) {
+		if (rhead->h_crc || xfs_sb_version_hascrc(&log->l_mp->m_sb)) {
+			xfs_alert(log->l_mp,
+		"log record CRC mismatch: found 0x%x, expected 0x%x.\n",
+					le32_to_cpu(rhead->h_crc),
+					le32_to_cpu(crc));
+			xfs_hex_dump(dp, 32);
 		}
 
-		for (i = 1; i < log->l_iclog_heads; i++) {
-			xhdr[i].hic_xheader.xh_cycle = cycle_lsn;
-		}
+		/*
+		 * If we've detected a log record corruption, then we can't
+		 * recover past this point. Abort recovery if we are enforcing
+		 * CRC protection by punting an error back up the stack.
+		 */
+		if (xfs_sb_version_hascrc(&log->l_mp->m_sb))
+			return EFSCORRUPTED;
 	}
+
+	return 0;
 }
 
-STATIC void
+STATIC int
 xlog_unpack_data(
 	struct xlog_rec_header	*rhead,
 	xfs_caddr_t		dp,
 	struct xlog		*log)
 {
 	int			i, j, k;
+	int			error;
+
+	error = xlog_unpack_data_crc(rhead, dp, log);
+	if (error)
+		return error;
 
 	for (i = 0; i < BTOBB(be32_to_cpu(rhead->h_len)) &&
 		  i < (XLOG_HEADER_CYCLE_SIZE / BBSIZE); i++) {
@@ -3303,6 +3285,8 @@ xlog_unpack_data(
 			dp += BBSIZE;
 		}
 	}
+
+	return 0;
 }
 
 STATIC int
@@ -3434,9 +3418,13 @@ xlog_do_recovery_pass(
 			if (error)
 				goto bread_err2;
 
-			xlog_unpack_data(rhead, offset, log);
-			if ((error = xlog_recover_process_data(log,
-						rhash, rhead, offset, pass)))
+			error = xlog_unpack_data(rhead, offset, log);
+			if (error)
+				goto bread_err2;
+
+			error = xlog_recover_process_data(log,
+						rhash, rhead, offset, pass);
+			if (error)
 				goto bread_err2;
 			blk_no += bblks + hblks;
 		}
@@ -3546,9 +3534,14 @@ xlog_do_recovery_pass(
 				if (error)
 					goto bread_err2;
 			}
-			xlog_unpack_data(rhead, offset, log);
-			if ((error = xlog_recover_process_data(log, rhash,
-							rhead, offset, pass)))
+
+			error = xlog_unpack_data(rhead, offset, log);
+			if (error)
+				goto bread_err2;
+
+			error = xlog_recover_process_data(log, rhash,
+							rhead, offset, pass);
+			if (error)
 				goto bread_err2;
 			blk_no += bblks;
 		}
@@ -3573,9 +3566,13 @@ xlog_do_recovery_pass(
 			if (error)
 				goto bread_err2;
 
-			xlog_unpack_data(rhead, offset, log);
-			if ((error = xlog_recover_process_data(log, rhash,
-							rhead, offset, pass)))
+			error = xlog_unpack_data(rhead, offset, log);
+			if (error)
+				goto bread_err2;
+
+			error = xlog_recover_process_data(log, rhash,
+							rhead, offset, pass);
+			if (error)
 				goto bread_err2;
 			blk_no += bblks + hblks;
 		}
@@ -3689,13 +3686,14 @@ xlog_do_recover(
 
 	/*
 	 * Now that we've finished replaying all buffer and inode
-	 * updates, re-read in the superblock.
+	 * updates, re-read in the superblock and reverify it.
 	 */
 	bp = xfs_getsb(log->l_mp, 0);
 	XFS_BUF_UNDONE(bp);
 	ASSERT(!(XFS_BUF_ISWRITE(bp)));
 	XFS_BUF_READ(bp);
 	XFS_BUF_UNASYNC(bp);
+	bp->b_ops = &xfs_sb_buf_ops;
 	xfsbdstrat(log->l_mp, bp);
 	error = xfs_buf_iowait(bp);
 	if (error) {
@@ -3707,7 +3705,7 @@ xlog_do_recover(
 
 	/* Convert superblock from on-disk format */
 	sbp = &log->l_mp->m_sb;
-	xfs_sb_from_disk(log->l_mp, XFS_BUF_TO_SBP(bp));
+	xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp));
 	ASSERT(sbp->sb_magicnum == XFS_SB_MAGIC);
 	ASSERT(xfs_sb_good_version(sbp));
 	xfs_buf_relse(bp);

fs/xfs/xfs_mount.c

@@ -42,6 +42,7 @@
 #include "xfs_fsops.h"
 #include "xfs_utils.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 
 #ifdef HAVE_PERCPU_SB
@@ -303,9 +304,8 @@ STATIC int
 xfs_mount_validate_sb(
 	xfs_mount_t	*mp,
 	xfs_sb_t	*sbp,
-	int		flags)
+	bool		check_inprogress)
 {
-	int		loud = !(flags & XFS_MFSI_QUIET);
 
 	/*
 	 * If the log device and data device have the
@@ -315,21 +315,18 @@ xfs_mount_validate_sb(
 	 * a volume filesystem in a non-volume manner.
 	 */
 	if (sbp->sb_magicnum != XFS_SB_MAGIC) {
-		if (loud)
-			xfs_warn(mp, "bad magic number");
+		xfs_warn(mp, "bad magic number");
 		return XFS_ERROR(EWRONGFS);
 	}
 
 	if (!xfs_sb_good_version(sbp)) {
-		if (loud)
-			xfs_warn(mp, "bad version");
+		xfs_warn(mp, "bad version");
 		return XFS_ERROR(EWRONGFS);
 	}
 
 	if (unlikely(
 	    sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
-		if (loud)
-			xfs_warn(mp,
+		xfs_warn(mp,
 		"filesystem is marked as having an external log; "
 		"specify logdev on the mount command line.");
 		return XFS_ERROR(EINVAL);
@@ -337,8 +334,7 @@ xfs_mount_validate_sb(
 
 	if (unlikely(
 	    sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
-		if (loud)
-			xfs_warn(mp,
+		xfs_warn(mp,
 		"filesystem is marked as having an internal log; "
 		"do not specify logdev on the mount command line.");
 		return XFS_ERROR(EINVAL);
@@ -372,8 +368,7 @@ xfs_mount_validate_sb(
 	    sbp->sb_dblocks == 0					||
 	    sbp->sb_dblocks > XFS_MAX_DBLOCKS(sbp)			||
 	    sbp->sb_dblocks < XFS_MIN_DBLOCKS(sbp))) {
-		if (loud)
-			XFS_CORRUPTION_ERROR("SB sanity check failed",
+		XFS_CORRUPTION_ERROR("SB sanity check failed",
 				XFS_ERRLEVEL_LOW, mp, sbp);
 		return XFS_ERROR(EFSCORRUPTED);
 	}
@@ -382,12 +377,10 @@ xfs_mount_validate_sb(
 	 * Until this is fixed only page-sized or smaller data blocks work.
 	 */
 	if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
-		if (loud) {
-			xfs_warn(mp,
+		xfs_warn(mp,
 		"File system with blocksize %d bytes. "
 		"Only pagesize (%ld) or less will currently work.",
 				sbp->sb_blocksize, PAGE_SIZE);
-		}
 		return XFS_ERROR(ENOSYS);
 	}
 
@@ -401,23 +394,20 @@ xfs_mount_validate_sb(
 	case 2048:
 		break;
 	default:
-		if (loud)
-			xfs_warn(mp, "inode size of %d bytes not supported",
+		xfs_warn(mp, "inode size of %d bytes not supported",
 				sbp->sb_inodesize);
 		return XFS_ERROR(ENOSYS);
 	}
 
 	if (xfs_sb_validate_fsb_count(sbp, sbp->sb_dblocks) ||
 	    xfs_sb_validate_fsb_count(sbp, sbp->sb_rblocks)) {
-		if (loud)
-			xfs_warn(mp,
+		xfs_warn(mp,
 		"file system too large to be mounted on this system.");
 		return XFS_ERROR(EFBIG);
 	}
 
-	if (unlikely(sbp->sb_inprogress)) {
-		if (loud)
-			xfs_warn(mp, "file system busy");
+	if (check_inprogress && sbp->sb_inprogress) {
+		xfs_warn(mp, "Offline file system operation in progress!");
 		return XFS_ERROR(EFSCORRUPTED);
 	}
 
@@ -425,9 +415,7 @@ xfs_mount_validate_sb(
 	 * Version 1 directory format has never worked on Linux.
 	 */
 	if (unlikely(!xfs_sb_version_hasdirv2(sbp))) {
-		if (loud)
-			xfs_warn(mp,
-				"file system using version 1 directory format");
+		xfs_warn(mp, "file system using version 1 directory format");
 		return XFS_ERROR(ENOSYS);
 	}
 
@@ -520,11 +508,9 @@ out_unwind:
 
 void
 xfs_sb_from_disk(
-	struct xfs_mount	*mp,
+	struct xfs_sb	*to,
 	xfs_dsb_t	*from)
 {
-	struct xfs_sb *to = &mp->m_sb;
-
 	to->sb_magicnum = be32_to_cpu(from->sb_magicnum);
 	to->sb_blocksize = be32_to_cpu(from->sb_blocksize);
 	to->sb_dblocks = be64_to_cpu(from->sb_dblocks);
@@ -626,6 +612,72 @@ xfs_sb_to_disk(
 	}
 }
 
+static void
+xfs_sb_verify(
+	struct xfs_buf	*bp)
+{
+	struct xfs_mount *mp = bp->b_target->bt_mount;
+	struct xfs_sb	sb;
+	int		error;
+
+	xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
+
+	/*
+	 * Only check the in progress field for the primary superblock as
+	 * mkfs.xfs doesn't clear it from secondary superblocks.
+	 */
+	error = xfs_mount_validate_sb(mp, &sb, bp->b_bn == XFS_SB_DADDR);
+	if (error)
+		xfs_buf_ioerror(bp, error);
+}
+
+static void
+xfs_sb_read_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_sb_verify(bp);
+}
+
+/*
+ * We may be probed for a filesystem match, so we may not want to emit
+ * messages when the superblock buffer is not actually an XFS superblock.
+ * If we find an XFS superblock, the run a normal, noisy mount because we are
+ * really going to mount it and want to know about errors.
+ */
+static void
+xfs_sb_quiet_read_verify(
+	struct xfs_buf	*bp)
+{
+	struct xfs_sb	sb;
+
+	xfs_sb_from_disk(&sb, XFS_BUF_TO_SBP(bp));
+
+	if (sb.sb_magicnum == XFS_SB_MAGIC) {
+		/* XFS filesystem, verify noisily! */
+		xfs_sb_read_verify(bp);
+		return;
+	}
+	/* quietly fail */
+	xfs_buf_ioerror(bp, EFSCORRUPTED);
+}
+
+static void
+xfs_sb_write_verify(
+	struct xfs_buf	*bp)
+{
+	xfs_sb_verify(bp);
+}
+
+const struct xfs_buf_ops xfs_sb_buf_ops = {
+	.verify_read = xfs_sb_read_verify,
+	.verify_write = xfs_sb_write_verify,
+};
+
+static const struct xfs_buf_ops xfs_sb_quiet_buf_ops = {
+	.verify_read = xfs_sb_quiet_read_verify,
+	.verify_write = xfs_sb_write_verify,
+};
+
 /*
  * xfs_readsb
  *
@@ -651,25 +703,26 @@ xfs_readsb(xfs_mount_t *mp, int flags)
 
 reread:
 	bp = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
-					BTOBB(sector_size), 0);
+				   BTOBB(sector_size), 0,
+				   loud ? &xfs_sb_buf_ops
+				        : &xfs_sb_quiet_buf_ops);
 	if (!bp) {
 		if (loud)
 			xfs_warn(mp, "SB buffer read failed");
 		return EIO;
 	}
-
-	/*
-	 * Initialize the mount structure from the superblock.
-	 * But first do some basic consistency checking.
-	 */
-	xfs_sb_from_disk(mp, XFS_BUF_TO_SBP(bp));
-	error = xfs_mount_validate_sb(mp, &(mp->m_sb), flags);
-	if (error) {
+	if (bp->b_error) {
+		error = bp->b_error;
 		if (loud)
 			xfs_warn(mp, "SB validate failed");
 		goto release_buf;
 	}
 
+	/*
+	 * Initialize the mount structure from the superblock.
+	 */
+	xfs_sb_from_disk(&mp->m_sb, XFS_BUF_TO_SBP(bp));
+
 	/*
 	 * We must be able to do sector-sized and sector-aligned IO.
 	 */
@@ -1001,7 +1054,7 @@ xfs_check_sizes(xfs_mount_t *mp)
 	}
 	bp = xfs_buf_read_uncached(mp->m_ddev_targp,
 					d - XFS_FSS_TO_BB(mp, 1),
-					XFS_FSS_TO_BB(mp, 1), 0);
+					XFS_FSS_TO_BB(mp, 1), 0, NULL);
 	if (!bp) {
 		xfs_warn(mp, "last sector read failed");
 		return EIO;
@@ -1016,7 +1069,7 @@ xfs_check_sizes(xfs_mount_t *mp)
 		}
 		bp = xfs_buf_read_uncached(mp->m_logdev_targp,
 					d - XFS_FSB_TO_BB(mp, 1),
-					XFS_FSB_TO_BB(mp, 1), 0);
+					XFS_FSB_TO_BB(mp, 1), 0, NULL);
 		if (!bp) {
 			xfs_warn(mp, "log device read failed");
 			return EIO;
@@ -1427,6 +1480,8 @@ xfs_unmountfs(
 	__uint64_t		resblks;
 	int			error;
 
+	cancel_delayed_work_sync(&mp->m_eofblocks_work);
+
 	xfs_qm_unmount_quotas(mp);
 	xfs_rtunmount_inodes(mp);
 	IRELE(mp->m_rootip);
@@ -1450,20 +1505,15 @@ xfs_unmountfs(
 
 	/*
 	 * And reclaim all inodes.  At this point there should be no dirty
-	 * inode, and none should be pinned or locked, but use synchronous
-	 * reclaim just to be sure.
+	 * inodes and none should be pinned or locked, but use synchronous
+	 * reclaim just to be sure. We can stop background inode reclaim
+	 * here as well if it is still running.
 	 */
+	cancel_delayed_work_sync(&mp->m_reclaim_work);
 	xfs_reclaim_inodes(mp, SYNC_WAIT);
 
 	xfs_qm_unmount(mp);
 
-	/*
-	 * Flush out the log synchronously so that we know for sure
-	 * that nothing is pinned.  This is important because bflush()
-	 * will skip pinned buffers.
-	 */
-	xfs_log_force(mp, XFS_LOG_SYNC);
-
 	/*
 	 * Unreserve any blocks we have so that when we unmount we don't account
 	 * the reserved free space as used. This is really only necessary for
@@ -1489,23 +1539,6 @@ xfs_unmountfs(
 		xfs_warn(mp, "Unable to update superblock counters. "
 				"Freespace may not be correct on next mount.");
 
-	/*
-	 * At this point we might have modified the superblock again and thus
-	 * added an item to the AIL, thus flush it again.
-	 */
-	xfs_ail_push_all_sync(mp->m_ail);
-	xfs_wait_buftarg(mp->m_ddev_targp);
-
-	/*
-	 * The superblock buffer is uncached and xfsaild_push() will lock and
-	 * set the XBF_ASYNC flag on the buffer. We cannot do xfs_buf_iowait()
-	 * here but a lock on the superblock buffer will block until iodone()
-	 * has completed.
-	 */
-	xfs_buf_lock(mp->m_sb_bp);
-	xfs_buf_unlock(mp->m_sb_bp);
-
-	xfs_log_unmount_write(mp);
 	xfs_log_unmount(mp);
 	xfs_uuid_unmount(mp);
 

fs/xfs/xfs_mount.h

@@ -51,8 +51,6 @@ typedef struct xfs_trans_reservations {
 
 #else /* __KERNEL__ */
 
-#include "xfs_sync.h"
-
 struct xlog;
 struct xfs_inode;
 struct xfs_mru_cache;
@@ -197,9 +195,9 @@ typedef struct xfs_mount {
 	struct mutex		m_icsb_mutex;	/* balancer sync lock */
 #endif
 	struct xfs_mru_cache	*m_filestream;  /* per-mount filestream data */
-	struct delayed_work	m_sync_work;	/* background sync work */
 	struct delayed_work	m_reclaim_work;	/* background inode reclaim */
-	struct work_struct	m_flush_work;	/* background inode flush */
+	struct delayed_work	m_eofblocks_work; /* background eof blocks
+						     trimming */
 	__int64_t		m_update_flags;	/* sb flags we need to update
 						   on the next remount,rw */
 	struct shrinker		m_inode_shrink;	/* inode reclaim shrinker */
@@ -209,6 +207,9 @@ typedef struct xfs_mount {
 	struct workqueue_struct	*m_data_workqueue;
 	struct workqueue_struct	*m_unwritten_workqueue;
 	struct workqueue_struct	*m_cil_workqueue;
+	struct workqueue_struct	*m_reclaim_workqueue;
+	struct workqueue_struct	*m_log_workqueue;
+	struct workqueue_struct *m_eofblocks_workqueue;
 } xfs_mount_t;
 
 /*
@@ -387,7 +388,9 @@ extern void	xfs_set_low_space_thresholds(struct xfs_mount *);
 extern void	xfs_mod_sb(struct xfs_trans *, __int64_t);
 extern int	xfs_initialize_perag(struct xfs_mount *, xfs_agnumber_t,
 					xfs_agnumber_t *);
-extern void	xfs_sb_from_disk(struct xfs_mount *, struct xfs_dsb *);
+extern void	xfs_sb_from_disk(struct xfs_sb *, struct xfs_dsb *);
 extern void	xfs_sb_to_disk(struct xfs_dsb *, struct xfs_sb *, __int64_t);
 
+extern const struct xfs_buf_ops xfs_sb_buf_ops;
+
 #endif	/* __XFS_MOUNT_H__ */

fs/xfs/xfs_qm.c

@@ -40,6 +40,7 @@
 #include "xfs_utils.h"
 #include "xfs_qm.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 /*
  * The global quota manager. There is only one of these for the entire
@@ -891,7 +892,8 @@ xfs_qm_dqiter_bufs(
 	while (blkcnt--) {
 		error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
 			      XFS_FSB_TO_DADDR(mp, bno),
-			      mp->m_quotainfo->qi_dqchunklen, 0, &bp);
+			      mp->m_quotainfo->qi_dqchunklen, 0, &bp,
+			      &xfs_dquot_buf_ops);
 		if (error)
 			break;
 
@@ -978,7 +980,8 @@ xfs_qm_dqiterate(
 				while (rablkcnt--) {
 					xfs_buf_readahead(mp->m_ddev_targp,
 					       XFS_FSB_TO_DADDR(mp, rablkno),
-					       mp->m_quotainfo->qi_dqchunklen);
+					       mp->m_quotainfo->qi_dqchunklen,
+					       NULL);
 					rablkno++;
 				}
 			}
@@ -1453,7 +1456,7 @@ xfs_qm_dqreclaim_one(
 	int			error;
 
 	if (!xfs_dqlock_nowait(dqp))
-		goto out_busy;
+		goto out_move_tail;
 
 	/*
 	 * This dquot has acquired a reference in the meantime remove it from
@@ -1476,7 +1479,7 @@ xfs_qm_dqreclaim_one(
 	 * getting flushed to disk, we don't want to reclaim it.
 	 */
 	if (!xfs_dqflock_nowait(dqp))
-		goto out_busy;
+		goto out_unlock_move_tail;
 
 	if (XFS_DQ_IS_DIRTY(dqp)) {
 		struct xfs_buf	*bp = NULL;
@@ -1487,7 +1490,7 @@ xfs_qm_dqreclaim_one(
 		if (error) {
 			xfs_warn(mp, "%s: dquot %p flush failed",
 				 __func__, dqp);
-			goto out_busy;
+			goto out_unlock_move_tail;
 		}
 
 		xfs_buf_delwri_queue(bp, buffer_list);
@@ -1496,7 +1499,7 @@ xfs_qm_dqreclaim_one(
 		 * Give the dquot another try on the freelist, as the
 		 * flushing will take some time.
 		 */
-		goto out_busy;
+		goto out_unlock_move_tail;
 	}
 	xfs_dqfunlock(dqp);
 
@@ -1515,14 +1518,13 @@ xfs_qm_dqreclaim_one(
 	XFS_STATS_INC(xs_qm_dqreclaims);
 	return;
 
-out_busy:
-	xfs_dqunlock(dqp);
-
 	/*
 	 * Move the dquot to the tail of the list so that we don't spin on it.
 	 */
+out_unlock_move_tail:
+	xfs_dqunlock(dqp);
+out_move_tail:
 	list_move_tail(&dqp->q_lru, &qi->qi_lru_list);
-
 	trace_xfs_dqreclaim_busy(dqp);
 	XFS_STATS_INC(xs_qm_dqreclaim_misses);
 }

fs/xfs/xfs_qm_syscalls.c

@@ -40,6 +40,7 @@
 #include "xfs_utils.h"
 #include "xfs_qm.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 STATIC int	xfs_qm_log_quotaoff(xfs_mount_t *, xfs_qoff_logitem_t **, uint);
 STATIC int	xfs_qm_log_quotaoff_end(xfs_mount_t *, xfs_qoff_logitem_t *,
@@ -845,7 +846,8 @@ STATIC int
 xfs_dqrele_inode(
 	struct xfs_inode	*ip,
 	struct xfs_perag	*pag,
-	int			flags)
+	int			flags,
+	void			*args)
 {
 	/* skip quota inodes */
 	if (ip == ip->i_mount->m_quotainfo->qi_uquotaip ||
@@ -881,5 +883,5 @@ xfs_qm_dqrele_all_inodes(
 	uint		 flags)
 {
 	ASSERT(mp->m_quotainfo);
-	xfs_inode_ag_iterator(mp, xfs_dqrele_inode, flags);
+	xfs_inode_ag_iterator(mp, xfs_dqrele_inode, flags, NULL);
 }

fs/xfs/xfs_rtalloc.c

@@ -38,6 +38,7 @@
 #include "xfs_utils.h"
 #include "xfs_trace.h"
 #include "xfs_buf.h"
+#include "xfs_icache.h"
 
 
 /*
@@ -869,7 +870,7 @@ xfs_rtbuf_get(
 	ASSERT(map.br_startblock != NULLFSBLOCK);
 	error = xfs_trans_read_buf(mp, tp, mp->m_ddev_targp,
 				   XFS_FSB_TO_DADDR(mp, map.br_startblock),
-				   mp->m_bsize, 0, &bp);
+				   mp->m_bsize, 0, &bp, NULL);
 	if (error)
 		return error;
 	ASSERT(!xfs_buf_geterror(bp));
@@ -1872,9 +1873,14 @@ xfs_growfs_rt(
 	 */
 	bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
 				XFS_FSB_TO_BB(mp, nrblocks - 1),
-				XFS_FSB_TO_BB(mp, 1), 0);
+				XFS_FSB_TO_BB(mp, 1), 0, NULL);
 	if (!bp)
 		return EIO;
+	if (bp->b_error) {
+		error = bp->b_error;
+		xfs_buf_relse(bp);
+		return error;
+	}
 	xfs_buf_relse(bp);
 
 	/*
@@ -2219,9 +2225,11 @@ xfs_rtmount_init(
 	}
 	bp = xfs_buf_read_uncached(mp->m_rtdev_targp,
 					d - XFS_FSB_TO_BB(mp, 1),
-					XFS_FSB_TO_BB(mp, 1), 0);
-	if (!bp) {
+					XFS_FSB_TO_BB(mp, 1), 0, NULL);
+	if (!bp || bp->b_error) {
 		xfs_warn(mp, "realtime device size check failed");
+		if (bp)
+			xfs_buf_relse(bp);
 		return EIO;
 	}
 	xfs_buf_relse(bp);

fs/xfs/xfs_sb.h

@@ -81,6 +81,7 @@ struct xfs_mount;
 #define XFS_SB_VERSION2_ATTR2BIT	0x00000008	/* Inline attr rework */
 #define XFS_SB_VERSION2_PARENTBIT	0x00000010	/* parent pointers */
 #define XFS_SB_VERSION2_PROJID32BIT	0x00000080	/* 32 bit project id */
+#define XFS_SB_VERSION2_CRCBIT		0x00000100	/* metadata CRCs */
 
 #define	XFS_SB_VERSION2_OKREALFBITS	\
 	(XFS_SB_VERSION2_LAZYSBCOUNTBIT	| \
@@ -503,6 +504,12 @@ static inline int xfs_sb_version_hasprojid32bit(xfs_sb_t *sbp)
 		(sbp->sb_features2 & XFS_SB_VERSION2_PROJID32BIT);
 }
 
+static inline int xfs_sb_version_hascrc(xfs_sb_t *sbp)
+{
+	return (xfs_sb_version_hasmorebits(sbp) &&
+		(sbp->sb_features2 & XFS_SB_VERSION2_CRCBIT));
+}
+
 /*
  * end of superblock version macros
  */

fs/xfs/xfs_super.c

@@ -49,7 +49,7 @@
 #include "xfs_extfree_item.h"
 #include "xfs_mru_cache.h"
 #include "xfs_inode_item.h"
-#include "xfs_sync.h"
+#include "xfs_icache.h"
 #include "xfs_trace.h"
 
 #include <linux/namei.h>
@@ -863,8 +863,30 @@ xfs_init_mount_workqueues(
 			WQ_MEM_RECLAIM, 0, mp->m_fsname);
 	if (!mp->m_cil_workqueue)
 		goto out_destroy_unwritten;
+
+	mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
+			WQ_NON_REENTRANT, 0, mp->m_fsname);
+	if (!mp->m_reclaim_workqueue)
+		goto out_destroy_cil;
+
+	mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
+			WQ_NON_REENTRANT, 0, mp->m_fsname);
+	if (!mp->m_log_workqueue)
+		goto out_destroy_reclaim;
+
+	mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
+			WQ_NON_REENTRANT, 0, mp->m_fsname);
+	if (!mp->m_eofblocks_workqueue)
+		goto out_destroy_log;
+
 	return 0;
 
+out_destroy_log:
+	destroy_workqueue(mp->m_log_workqueue);
+out_destroy_reclaim:
+	destroy_workqueue(mp->m_reclaim_workqueue);
+out_destroy_cil:
+	destroy_workqueue(mp->m_cil_workqueue);
 out_destroy_unwritten:
 	destroy_workqueue(mp->m_unwritten_workqueue);
 out_destroy_data_iodone_queue:
@@ -877,11 +899,32 @@ STATIC void
 xfs_destroy_mount_workqueues(
 	struct xfs_mount	*mp)
 {
+	destroy_workqueue(mp->m_eofblocks_workqueue);
+	destroy_workqueue(mp->m_log_workqueue);
+	destroy_workqueue(mp->m_reclaim_workqueue);
 	destroy_workqueue(mp->m_cil_workqueue);
 	destroy_workqueue(mp->m_data_workqueue);
 	destroy_workqueue(mp->m_unwritten_workqueue);
 }
 
+/*
+ * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
+ * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
+ * for IO to complete so that we effectively throttle multiple callers to the
+ * rate at which IO is completing.
+ */
+void
+xfs_flush_inodes(
+	struct xfs_mount	*mp)
+{
+	struct super_block	*sb = mp->m_super;
+
+	if (down_read_trylock(&sb->s_umount)) {
+		sync_inodes_sb(sb);
+		up_read(&sb->s_umount);
+	}
+}
+
 /* Catch misguided souls that try to use this interface on XFS */
 STATIC struct inode *
 xfs_fs_alloc_inode(
@@ -1006,9 +1049,8 @@ xfs_fs_put_super(
 	struct xfs_mount	*mp = XFS_M(sb);
 
 	xfs_filestream_unmount(mp);
-	cancel_delayed_work_sync(&mp->m_sync_work);
 	xfs_unmountfs(mp);
-	xfs_syncd_stop(mp);
+
 	xfs_freesb(mp);
 	xfs_icsb_destroy_counters(mp);
 	xfs_destroy_mount_workqueues(mp);
@@ -1023,7 +1065,6 @@ xfs_fs_sync_fs(
 	int			wait)
 {
 	struct xfs_mount	*mp = XFS_M(sb);
-	int			error;
 
 	/*
 	 * Doing anything during the async pass would be counterproductive.
@@ -1031,17 +1072,14 @@ xfs_fs_sync_fs(
 	if (!wait)
 		return 0;
 
-	error = xfs_quiesce_data(mp);
-	if (error)
-		return -error;
-
+	xfs_log_force(mp, XFS_LOG_SYNC);
 	if (laptop_mode) {
 		/*
 		 * The disk must be active because we're syncing.
-		 * We schedule xfssyncd now (now that the disk is
+		 * We schedule log work now (now that the disk is
 		 * active) instead of later (when it might not be).
 		 */
-		flush_delayed_work(&mp->m_sync_work);
+		flush_delayed_work(&mp->m_log->l_work);
 	}
 
 	return 0;
@@ -1118,6 +1156,48 @@ xfs_restore_resvblks(struct xfs_mount *mp)
 	xfs_reserve_blocks(mp, &resblks, NULL);
 }
 
+/*
+ * Trigger writeback of all the dirty metadata in the file system.
+ *
+ * This ensures that the metadata is written to their location on disk rather
+ * than just existing in transactions in the log. This means after a quiesce
+ * there is no log replay required to write the inodes to disk - this is the
+ * primary difference between a sync and a quiesce.
+ *
+ * Note: xfs_log_quiesce() stops background log work - the callers must ensure
+ * it is started again when appropriate.
+ */
+void
+xfs_quiesce_attr(
+	struct xfs_mount	*mp)
+{
+	int	error = 0;
+
+	/* wait for all modifications to complete */
+	while (atomic_read(&mp->m_active_trans) > 0)
+		delay(100);
+
+	/* force the log to unpin objects from the now complete transactions */
+	xfs_log_force(mp, XFS_LOG_SYNC);
+
+	/* reclaim inodes to do any IO before the freeze completes */
+	xfs_reclaim_inodes(mp, 0);
+	xfs_reclaim_inodes(mp, SYNC_WAIT);
+
+	/* Push the superblock and write an unmount record */
+	error = xfs_log_sbcount(mp);
+	if (error)
+		xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
+				"Frozen image may not be consistent.");
+	/*
+	 * Just warn here till VFS can correctly support
+	 * read-only remount without racing.
+	 */
+	WARN_ON(atomic_read(&mp->m_active_trans) != 0);
+
+	xfs_log_quiesce(mp);
+}
+
 STATIC int
 xfs_fs_remount(
 	struct super_block	*sb,
@@ -1198,20 +1278,18 @@ xfs_fs_remount(
 		 * value if it is non-zero, otherwise go with the default.
 		 */
 		xfs_restore_resvblks(mp);
+		xfs_log_work_queue(mp);
 	}
 
 	/* rw -> ro */
 	if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
 		/*
-		 * After we have synced the data but before we sync the
-		 * metadata, we need to free up the reserve block pool so that
-		 * the used block count in the superblock on disk is correct at
-		 * the end of the remount. Stash the current reserve pool size
-		 * so that if we get remounted rw, we can return it to the same
-		 * size.
+		 * Before we sync the metadata, we need to free up the reserve
+		 * block pool so that the used block count in the superblock on
+		 * disk is correct at the end of the remount. Stash the current
+		 * reserve pool size so that if we get remounted rw, we can
+		 * return it to the same size.
 		 */
-
-		xfs_quiesce_data(mp);
 		xfs_save_resvblks(mp);
 		xfs_quiesce_attr(mp);
 		mp->m_flags |= XFS_MOUNT_RDONLY;
@@ -1243,6 +1321,7 @@ xfs_fs_unfreeze(
 	struct xfs_mount	*mp = XFS_M(sb);
 
 	xfs_restore_resvblks(mp);
+	xfs_log_work_queue(mp);
 	return 0;
 }
 
@@ -1321,6 +1400,8 @@ xfs_fs_fill_super(
 	spin_lock_init(&mp->m_sb_lock);
 	mutex_init(&mp->m_growlock);
 	atomic_set(&mp->m_active_trans, 0);
+	INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
+	INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
 
 	mp->m_super = sb;
 	sb->s_fs_info = mp;
@@ -1371,10 +1452,6 @@ xfs_fs_fill_super(
 	/*
 	 * we must configure the block size in the superblock before we run the
 	 * full mount process as the mount process can lookup and cache inodes.
-	 * For the same reason we must also initialise the syncd and register
-	 * the inode cache shrinker so that inodes can be reclaimed during
-	 * operations like a quotacheck that iterate all inodes in the
-	 * filesystem.
 	 */
 	sb->s_magic = XFS_SB_MAGIC;
 	sb->s_blocksize = mp->m_sb.sb_blocksize;
@@ -1384,13 +1461,9 @@ xfs_fs_fill_super(
 	sb->s_time_gran = 1;
 	set_posix_acl_flag(sb);
 
-	error = xfs_syncd_init(mp);
-	if (error)
-		goto out_filestream_unmount;
-
 	error = xfs_mountfs(mp);
 	if (error)
-		goto out_syncd_stop;
+		goto out_filestream_unmount;
 
 	root = igrab(VFS_I(mp->m_rootip));
 	if (!root) {
@@ -1408,8 +1481,7 @@ xfs_fs_fill_super(
 	}
 
 	return 0;
- out_syncd_stop:
-	xfs_syncd_stop(mp);
+
  out_filestream_unmount:
 	xfs_filestream_unmount(mp);
  out_free_sb:
@@ -1429,7 +1501,6 @@ out_destroy_workqueues:
  out_unmount:
 	xfs_filestream_unmount(mp);
 	xfs_unmountfs(mp);
-	xfs_syncd_stop(mp);
 	goto out_free_sb;
 }
 
@@ -1624,16 +1695,6 @@ xfs_destroy_zones(void)
 STATIC int __init
 xfs_init_workqueues(void)
 {
-	/*
-	 * We never want to the same work item to run twice, reclaiming inodes
-	 * or idling the log is not going to get any faster by multiple CPUs
-	 * competing for ressources.  Use the default large max_active value
-	 * so that even lots of filesystems can perform these task in parallel.
-	 */
-	xfs_syncd_wq = alloc_workqueue("xfssyncd", WQ_NON_REENTRANT, 0);
-	if (!xfs_syncd_wq)
-		return -ENOMEM;
-
 	/*
 	 * The allocation workqueue can be used in memory reclaim situations
 	 * (writepage path), and parallelism is only limited by the number of
@@ -1642,20 +1703,15 @@ xfs_init_workqueues(void)
 	 */
 	xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
 	if (!xfs_alloc_wq)
-		goto out_destroy_syncd;
+		return -ENOMEM;
 
 	return 0;
-
-out_destroy_syncd:
-	destroy_workqueue(xfs_syncd_wq);
-	return -ENOMEM;
 }
 
 STATIC void
 xfs_destroy_workqueues(void)
 {
 	destroy_workqueue(xfs_alloc_wq);
-	destroy_workqueue(xfs_syncd_wq);
 }
 
 STATIC int __init

fs/xfs/xfs_super.h

@@ -74,6 +74,7 @@ struct block_device;
 
 extern __uint64_t xfs_max_file_offset(unsigned int);
 
+extern void xfs_flush_inodes(struct xfs_mount *mp);
 extern void xfs_blkdev_issue_flush(struct xfs_buftarg *);
 extern xfs_agnumber_t xfs_set_inode32(struct xfs_mount *);
 extern xfs_agnumber_t xfs_set_inode64(struct xfs_mount *);

fs/xfs/xfs_sysctl.c

@@ -202,6 +202,15 @@ static ctl_table xfs_table[] = {
 		.extra1		= &xfs_params.fstrm_timer.min,
 		.extra2		= &xfs_params.fstrm_timer.max,
 	},
+	{
+		.procname	= "speculative_prealloc_lifetime",
+		.data		= &xfs_params.eofb_timer.val,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &xfs_params.eofb_timer.min,
+		.extra2		= &xfs_params.eofb_timer.max,
+	},
 	/* please keep this the last entry */
 #ifdef CONFIG_PROC_FS
 	{

fs/xfs/xfs_sysctl.h

@@ -47,6 +47,7 @@ typedef struct xfs_param {
 	xfs_sysctl_val_t rotorstep;	/* inode32 AG rotoring control knob */
 	xfs_sysctl_val_t inherit_nodfrg;/* Inherit the "nodefrag" inode flag. */
 	xfs_sysctl_val_t fstrm_timer;	/* Filestream dir-AG assoc'n timeout. */
+	xfs_sysctl_val_t eofb_timer;	/* Interval between eofb scan wakeups */
 } xfs_param_t;
 
 /*

fs/xfs/xfs_trace.h

@@ -96,6 +96,8 @@ DEFINE_ATTR_LIST_EVENT(xfs_attr_list_full);
 DEFINE_ATTR_LIST_EVENT(xfs_attr_list_add);
 DEFINE_ATTR_LIST_EVENT(xfs_attr_list_wrong_blk);
 DEFINE_ATTR_LIST_EVENT(xfs_attr_list_notfound);
+DEFINE_ATTR_LIST_EVENT(xfs_attr_leaf_list);
+DEFINE_ATTR_LIST_EVENT(xfs_attr_node_list);
 
 DECLARE_EVENT_CLASS(xfs_perag_class,
 	TP_PROTO(struct xfs_mount *mp, xfs_agnumber_t agno, int refcount,
@@ -130,6 +132,8 @@ DEFINE_PERAG_REF_EVENT(xfs_perag_get_tag);
 DEFINE_PERAG_REF_EVENT(xfs_perag_put);
 DEFINE_PERAG_REF_EVENT(xfs_perag_set_reclaim);
 DEFINE_PERAG_REF_EVENT(xfs_perag_clear_reclaim);
+DEFINE_PERAG_REF_EVENT(xfs_perag_set_eofblocks);
+DEFINE_PERAG_REF_EVENT(xfs_perag_clear_eofblocks);
 
 TRACE_EVENT(xfs_attr_list_node_descend,
 	TP_PROTO(struct xfs_attr_list_context *ctx,
@@ -585,6 +589,10 @@ DEFINE_INODE_EVENT(xfs_update_time);
 DEFINE_INODE_EVENT(xfs_dquot_dqalloc);
 DEFINE_INODE_EVENT(xfs_dquot_dqdetach);
 
+DEFINE_INODE_EVENT(xfs_inode_set_eofblocks_tag);
+DEFINE_INODE_EVENT(xfs_inode_clear_eofblocks_tag);
+DEFINE_INODE_EVENT(xfs_inode_free_eofblocks_invalid);
+
 DECLARE_EVENT_CLASS(xfs_iref_class,
 	TP_PROTO(struct xfs_inode *ip, unsigned long caller_ip),
 	TP_ARGS(ip, caller_ip),
@@ -1496,8 +1504,42 @@ DEFINE_DIR2_EVENT(xfs_dir2_node_replace);
 DEFINE_DIR2_EVENT(xfs_dir2_node_removename);
 DEFINE_DIR2_EVENT(xfs_dir2_node_to_leaf);
 
+DECLARE_EVENT_CLASS(xfs_attr_class,
+	TP_PROTO(struct xfs_da_args *args),
+	TP_ARGS(args),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(xfs_ino_t, ino)
+		__dynamic_array(char, name, args->namelen)
+		__field(int, namelen)
+		__field(int, valuelen)
+		__field(xfs_dahash_t, hashval)
+		__field(int, op_flags)
+	),
+	TP_fast_assign(
+		__entry->dev = VFS_I(args->dp)->i_sb->s_dev;
+		__entry->ino = args->dp->i_ino;
+		if (args->namelen)
+			memcpy(__get_str(name), args->name, args->namelen);
+		__entry->namelen = args->namelen;
+		__entry->valuelen = args->valuelen;
+		__entry->hashval = args->hashval;
+		__entry->op_flags = args->op_flags;
+	),
+	TP_printk("dev %d:%d ino 0x%llx name %.*s namelen %d valuelen %d "
+		  "hashval 0x%x op_flags %s",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __entry->ino,
+		  __entry->namelen,
+		  __entry->namelen ? __get_str(name) : NULL,
+		  __entry->namelen,
+		  __entry->valuelen,
+		  __entry->hashval,
+		  __print_flags(__entry->op_flags, "|", XFS_DA_OP_FLAGS))
+)
+
 #define DEFINE_ATTR_EVENT(name) \
-DEFINE_EVENT(xfs_da_class, name, \
+DEFINE_EVENT(xfs_attr_class, name, \
 	TP_PROTO(struct xfs_da_args *args), \
 	TP_ARGS(args))
 DEFINE_ATTR_EVENT(xfs_attr_sf_add);
@@ -1511,10 +1553,14 @@ DEFINE_ATTR_EVENT(xfs_attr_sf_to_leaf);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_add);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_add_old);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_add_new);
+DEFINE_ATTR_EVENT(xfs_attr_leaf_add_work);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_addname);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_create);
+DEFINE_ATTR_EVENT(xfs_attr_leaf_compact);
+DEFINE_ATTR_EVENT(xfs_attr_leaf_get);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_lookup);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_replace);
+DEFINE_ATTR_EVENT(xfs_attr_leaf_remove);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_removename);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_split);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_split_before);
@@ -1526,12 +1572,21 @@ DEFINE_ATTR_EVENT(xfs_attr_leaf_to_sf);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_to_node);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_rebalance);
 DEFINE_ATTR_EVENT(xfs_attr_leaf_unbalance);
+DEFINE_ATTR_EVENT(xfs_attr_leaf_toosmall);
 
 DEFINE_ATTR_EVENT(xfs_attr_node_addname);
+DEFINE_ATTR_EVENT(xfs_attr_node_get);
 DEFINE_ATTR_EVENT(xfs_attr_node_lookup);
 DEFINE_ATTR_EVENT(xfs_attr_node_replace);
 DEFINE_ATTR_EVENT(xfs_attr_node_removename);
 
+DEFINE_ATTR_EVENT(xfs_attr_fillstate);
+DEFINE_ATTR_EVENT(xfs_attr_refillstate);
+
+DEFINE_ATTR_EVENT(xfs_attr_rmtval_get);
+DEFINE_ATTR_EVENT(xfs_attr_rmtval_set);
+DEFINE_ATTR_EVENT(xfs_attr_rmtval_remove);
+
 #define DEFINE_DA_EVENT(name) \
 DEFINE_EVENT(xfs_da_class, name, \
 	TP_PROTO(struct xfs_da_args *args), \
@@ -1550,9 +1605,12 @@ DEFINE_DA_EVENT(xfs_da_node_split);
 DEFINE_DA_EVENT(xfs_da_node_remove);
 DEFINE_DA_EVENT(xfs_da_node_rebalance);
 DEFINE_DA_EVENT(xfs_da_node_unbalance);
+DEFINE_DA_EVENT(xfs_da_node_toosmall);
 DEFINE_DA_EVENT(xfs_da_swap_lastblock);
 DEFINE_DA_EVENT(xfs_da_grow_inode);
 DEFINE_DA_EVENT(xfs_da_shrink_inode);
+DEFINE_DA_EVENT(xfs_da_fixhashpath);
+DEFINE_DA_EVENT(xfs_da_path_shift);
 
 DECLARE_EVENT_CLASS(xfs_dir2_space_class,
 	TP_PROTO(struct xfs_da_args *args, int idx),

fs/xfs/xfs_trans.h

@@ -464,10 +464,7 @@ xfs_trans_get_buf(
 	int			numblks,
 	uint			flags)
 {
-	struct xfs_buf_map	map = {
-		.bm_bn = blkno,
-		.bm_len = numblks,
-	};
+	DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
 	return xfs_trans_get_buf_map(tp, target, &map, 1, flags);
 }
 
@@ -476,7 +473,8 @@ int		xfs_trans_read_buf_map(struct xfs_mount *mp,
 				       struct xfs_buftarg *target,
 				       struct xfs_buf_map *map, int nmaps,
 				       xfs_buf_flags_t flags,
-				       struct xfs_buf **bpp);
+				       struct xfs_buf **bpp,
+				       const struct xfs_buf_ops *ops);
 
 static inline int
 xfs_trans_read_buf(
@@ -486,13 +484,12 @@ xfs_trans_read_buf(
 	xfs_daddr_t		blkno,
 	int			numblks,
 	xfs_buf_flags_t		flags,
-	struct xfs_buf		**bpp)
+	struct xfs_buf		**bpp,
+	const struct xfs_buf_ops *ops)
 {
-	struct xfs_buf_map	map = {
-		.bm_bn = blkno,
-		.bm_len = numblks,
-	};
-	return xfs_trans_read_buf_map(mp, tp, target, &map, 1, flags, bpp);
+	DEFINE_SINGLE_BUF_MAP(map, blkno, numblks);
+	return xfs_trans_read_buf_map(mp, tp, target, &map, 1,
+				      flags, bpp, ops);
 }
 
 struct xfs_buf	*xfs_trans_getsb(xfs_trans_t *, struct xfs_mount *, int);

fs/xfs/xfs_trans_buf.c

@@ -257,7 +257,8 @@ xfs_trans_read_buf_map(
 	struct xfs_buf_map	*map,
 	int			nmaps,
 	xfs_buf_flags_t		flags,
-	struct xfs_buf		**bpp)
+	struct xfs_buf		**bpp,
+	const struct xfs_buf_ops *ops)
 {
 	xfs_buf_t		*bp;
 	xfs_buf_log_item_t	*bip;
@@ -265,7 +266,7 @@ xfs_trans_read_buf_map(
 
 	*bpp = NULL;
 	if (!tp) {
-		bp = xfs_buf_read_map(target, map, nmaps, flags);
+		bp = xfs_buf_read_map(target, map, nmaps, flags, ops);
 		if (!bp)
 			return (flags & XBF_TRYLOCK) ?
 					EAGAIN : XFS_ERROR(ENOMEM);
@@ -312,7 +313,9 @@ xfs_trans_read_buf_map(
 		if (!(XFS_BUF_ISDONE(bp))) {
 			trace_xfs_trans_read_buf_io(bp, _RET_IP_);
 			ASSERT(!XFS_BUF_ISASYNC(bp));
+			ASSERT(bp->b_iodone == NULL);
 			XFS_BUF_READ(bp);
+			bp->b_ops = ops;
 			xfsbdstrat(tp->t_mountp, bp);
 			error = xfs_buf_iowait(bp);
 			if (error) {
@@ -349,7 +352,7 @@ xfs_trans_read_buf_map(
 		return 0;
 	}
 
-	bp = xfs_buf_read_map(target, map, nmaps, flags);
+	bp = xfs_buf_read_map(target, map, nmaps, flags, ops);
 	if (bp == NULL) {
 		*bpp = NULL;
 		return (flags & XBF_TRYLOCK) ?

fs/xfs/xfs_vnodeops.c

@@ -47,6 +47,7 @@
 #include "xfs_filestream.h"
 #include "xfs_vnodeops.h"
 #include "xfs_trace.h"
+#include "xfs_icache.h"
 
 /*
  * The maximum pathlen is 1024 bytes. Since the minimum file system
@@ -79,7 +80,7 @@ xfs_readlink_bmap(
 		d = XFS_FSB_TO_DADDR(mp, mval[n].br_startblock);
 		byte_cnt = XFS_FSB_TO_B(mp, mval[n].br_blockcount);
 
-		bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0);
+		bp = xfs_buf_read(mp->m_ddev_targp, d, BTOBB(byte_cnt), 0, NULL);
 		if (!bp)
 			return XFS_ERROR(ENOMEM);
 		error = bp->b_error;
@@ -150,7 +151,7 @@ xfs_readlink(
  * when the link count isn't zero and by xfs_dm_punch_hole() when
  * punching a hole to EOF.
  */
-STATIC int
+int
 xfs_free_eofblocks(
 	xfs_mount_t	*mp,
 	xfs_inode_t	*ip,
@@ -199,7 +200,7 @@ xfs_free_eofblocks(
 		if (need_iolock) {
 			if (!xfs_ilock_nowait(ip, XFS_IOLOCK_EXCL)) {
 				xfs_trans_cancel(tp, 0);
-				return 0;
+				return EAGAIN;
 			}
 		}
 
@@ -237,6 +238,8 @@ xfs_free_eofblocks(
 		} else {
 			error = xfs_trans_commit(tp,
 						XFS_TRANS_RELEASE_LOG_RES);
+			if (!error)
+				xfs_inode_clear_eofblocks_tag(ip);
 		}
 
 		xfs_iunlock(ip, XFS_ILOCK_EXCL);
@@ -425,19 +428,18 @@ xfs_release(
 		truncated = xfs_iflags_test_and_clear(ip, XFS_ITRUNCATED);
 		if (truncated) {
 			xfs_iflags_clear(ip, XFS_IDIRTY_RELEASE);
-			if (VN_DIRTY(VFS_I(ip)) && ip->i_delayed_blks > 0)
-				xfs_flush_pages(ip, 0, -1, XBF_ASYNC, FI_NONE);
+			if (VN_DIRTY(VFS_I(ip)) && ip->i_delayed_blks > 0) {
+				error = -filemap_flush(VFS_I(ip)->i_mapping);
+				if (error)
+					return error;
+			}
 		}
 	}
 
 	if (ip->i_d.di_nlink == 0)
 		return 0;
 
-	if ((S_ISREG(ip->i_d.di_mode) &&
-	     (VFS_I(ip)->i_size > 0 ||
-	      (VN_CACHED(VFS_I(ip)) > 0 || ip->i_delayed_blks > 0)) &&
-	     (ip->i_df.if_flags & XFS_IFEXTENTS))  &&
-	    (!(ip->i_d.di_flags & (XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)))) {
+	if (xfs_can_free_eofblocks(ip, false)) {
 
 		/*
 		 * If we can't get the iolock just skip truncating the blocks
@@ -464,7 +466,7 @@ xfs_release(
 			return 0;
 
 		error = xfs_free_eofblocks(mp, ip, true);
-		if (error)
+		if (error && error != EAGAIN)
 			return error;
 
 		/* delalloc blocks after truncation means it really is dirty */
@@ -513,13 +515,12 @@ xfs_inactive(
 		goto out;
 
 	if (ip->i_d.di_nlink != 0) {
-		if ((S_ISREG(ip->i_d.di_mode) &&
-		    (VFS_I(ip)->i_size > 0 ||
-		     (VN_CACHED(VFS_I(ip)) > 0 || ip->i_delayed_blks > 0)) &&
-		    (ip->i_df.if_flags & XFS_IFEXTENTS) &&
-		    (!(ip->i_d.di_flags &
-				(XFS_DIFLAG_PREALLOC | XFS_DIFLAG_APPEND)) ||
-		     ip->i_delayed_blks != 0))) {
+		/*
+		 * force is true because we are evicting an inode from the
+		 * cache. Post-eof blocks must be freed, lest we end up with
+		 * broken free space accounting.
+		 */
+		if (xfs_can_free_eofblocks(ip, true)) {
 			error = xfs_free_eofblocks(mp, ip, false);
 			if (error)
 				return VN_INACTIVE_CACHE;
@@ -777,7 +778,7 @@ xfs_create(
 			XFS_TRANS_PERM_LOG_RES, log_count);
 	if (error == ENOSPC) {
 		/* flush outstanding delalloc blocks and retry */
-		xfs_flush_inodes(dp);
+		xfs_flush_inodes(mp);
 		error = xfs_trans_reserve(tp, resblks, log_res, 0,
 				XFS_TRANS_PERM_LOG_RES, log_count);
 	}
@@ -1957,12 +1958,11 @@ xfs_free_file_space(
 
 	rounding = max_t(uint, 1 << mp->m_sb.sb_blocklog, PAGE_CACHE_SIZE);
 	ioffset = offset & ~(rounding - 1);
-
-	if (VN_CACHED(VFS_I(ip)) != 0) {
-		error = xfs_flushinval_pages(ip, ioffset, -1, FI_REMAPF_LOCKED);
-		if (error)
-			goto out_unlock_iolock;
-	}
+	error = -filemap_write_and_wait_range(VFS_I(ip)->i_mapping,
+					      ioffset, -1);
+	if (error)
+		goto out_unlock_iolock;
+	truncate_pagecache_range(VFS_I(ip), ioffset, -1);
 
 	/*
 	 * Need to zero the stuff we're not freeing, on disk.
@@ -2095,6 +2095,73 @@ xfs_free_file_space(
 	return error;
 }
 
+
+STATIC int
+xfs_zero_file_space(
+	struct xfs_inode	*ip,
+	xfs_off_t		offset,
+	xfs_off_t		len,
+	int			attr_flags)
+{
+	struct xfs_mount	*mp = ip->i_mount;
+	uint			granularity;
+	xfs_off_t		start_boundary;
+	xfs_off_t		end_boundary;
+	int			error;
+
+	granularity = max_t(uint, 1 << mp->m_sb.sb_blocklog, PAGE_CACHE_SIZE);
+
+	/*
+	 * Round the range of extents we are going to convert inwards.  If the
+	 * offset is aligned, then it doesn't get changed so we zero from the
+	 * start of the block offset points to.
+	 */
+	start_boundary = round_up(offset, granularity);
+	end_boundary = round_down(offset + len, granularity);
+
+	ASSERT(start_boundary >= offset);
+	ASSERT(end_boundary <= offset + len);
+
+	if (!(attr_flags & XFS_ATTR_NOLOCK))
+		xfs_ilock(ip, XFS_IOLOCK_EXCL);
+
+	if (start_boundary < end_boundary - 1) {
+		/* punch out the page cache over the conversion range */
+		truncate_pagecache_range(VFS_I(ip), start_boundary,
+					 end_boundary - 1);
+		/* convert the blocks */
+		error = xfs_alloc_file_space(ip, start_boundary,
+					end_boundary - start_boundary - 1,
+					XFS_BMAPI_PREALLOC | XFS_BMAPI_CONVERT,
+					attr_flags);
+		if (error)
+			goto out_unlock;
+
+		/* We've handled the interior of the range, now for the edges */
+		if (start_boundary != offset)
+			error = xfs_iozero(ip, offset, start_boundary - offset);
+		if (error)
+			goto out_unlock;
+
+		if (end_boundary != offset + len)
+			error = xfs_iozero(ip, end_boundary,
+					   offset + len - end_boundary);
+
+	} else {
+		/*
+		 * It's either a sub-granularity range or the range spanned lies
+		 * partially across two adjacent blocks.
+		 */
+		error = xfs_iozero(ip, offset, len);
+	}
+
+out_unlock:
+	if (!(attr_flags & XFS_ATTR_NOLOCK))
+		xfs_iunlock(ip, XFS_IOLOCK_EXCL);
+	return error;
+
+}
+
 /*
  * xfs_change_file_space()
  *      This routine allocates or frees disk space for the given file.
@@ -2120,10 +2187,8 @@ xfs_change_file_space(
 	xfs_fsize_t	fsize;
 	int		setprealloc;
 	xfs_off_t	startoffset;
-	xfs_off_t	llen;
 	xfs_trans_t	*tp;
 	struct iattr	iattr;
-	int		prealloc_type;
 
 	if (!S_ISREG(ip->i_d.di_mode))
 		return XFS_ERROR(EINVAL);
@@ -2141,12 +2206,30 @@ xfs_change_file_space(
 		return XFS_ERROR(EINVAL);
 	}
 
-	llen = bf->l_len > 0 ? bf->l_len - 1 : bf->l_len;
+	/*
+	 * length of <= 0 for resv/unresv/zero is invalid.  length for
+	 * alloc/free is ignored completely and we have no idea what userspace
+	 * might have set it to, so set it to zero to allow range
+	 * checks to pass.
+	 */
+	switch (cmd) {
+	case XFS_IOC_ZERO_RANGE:
+	case XFS_IOC_RESVSP:
+	case XFS_IOC_RESVSP64:
+	case XFS_IOC_UNRESVSP:
+	case XFS_IOC_UNRESVSP64:
+		if (bf->l_len <= 0)
+			return XFS_ERROR(EINVAL);
+		break;
+	default:
+		bf->l_len = 0;
+		break;
+	}
 
 	if (bf->l_start < 0 ||
 	    bf->l_start > mp->m_super->s_maxbytes ||
-	    bf->l_start + llen < 0 ||
-	    bf->l_start + llen > mp->m_super->s_maxbytes)
+	    bf->l_start + bf->l_len < 0 ||
+	    bf->l_start + bf->l_len >= mp->m_super->s_maxbytes)
 		return XFS_ERROR(EINVAL);
 
 	bf->l_whence = 0;
@@ -2154,29 +2237,20 @@ xfs_change_file_space(
 	startoffset = bf->l_start;
 	fsize = XFS_ISIZE(ip);
 
-	/*
-	 * XFS_IOC_RESVSP and XFS_IOC_UNRESVSP will reserve or unreserve
-	 * file space.
-	 * These calls do NOT zero the data space allocated to the file,
-	 * nor do they change the file size.
-	 *
-	 * XFS_IOC_ALLOCSP and XFS_IOC_FREESP will allocate and free file
-	 * space.
-	 * These calls cause the new file data to be zeroed and the file
-	 * size to be changed.
-	 */
 	setprealloc = clrprealloc = 0;
-	prealloc_type = XFS_BMAPI_PREALLOC;
-
 	switch (cmd) {
 	case XFS_IOC_ZERO_RANGE:
-		prealloc_type |= XFS_BMAPI_CONVERT;
-		xfs_tosspages(ip, startoffset, startoffset + bf->l_len, 0);
-		/* FALLTHRU */
+		error = xfs_zero_file_space(ip, startoffset, bf->l_len,
+						attr_flags);
+		if (error)
+			return error;
+		setprealloc = 1;
+		break;
+
 	case XFS_IOC_RESVSP:
 	case XFS_IOC_RESVSP64:
 		error = xfs_alloc_file_space(ip, startoffset, bf->l_len,
-						prealloc_type, attr_flags);
+						XFS_BMAPI_PREALLOC, attr_flags);
 		if (error)
 			return error;
 		setprealloc = 1;

fs/xfs/xfs_vnodeops.h

@@ -48,14 +48,9 @@ int xfs_attr_set(struct xfs_inode *dp, const unsigned char *name,
 int xfs_attr_remove(struct xfs_inode *dp, const unsigned char *name, int flags);
 int xfs_attr_list(struct xfs_inode *dp, char *buffer, int bufsize,
 		int flags, struct attrlist_cursor_kern *cursor);
-void xfs_tosspages(struct xfs_inode *inode, xfs_off_t first,
-		xfs_off_t last, int fiopt);
-int xfs_flushinval_pages(struct xfs_inode *ip, xfs_off_t first,
-		xfs_off_t last, int fiopt);
-int xfs_flush_pages(struct xfs_inode *ip, xfs_off_t first,
-		xfs_off_t last, uint64_t flags, int fiopt);
-int xfs_wait_on_pages(struct xfs_inode *ip, xfs_off_t first, xfs_off_t last);
 
+int xfs_iozero(struct xfs_inode *, loff_t, size_t);
 int xfs_zero_eof(struct xfs_inode *, xfs_off_t, xfs_fsize_t);
+int xfs_free_eofblocks(struct xfs_mount *, struct xfs_inode *, bool);
 
 #endif /* _XFS_VNODEOPS_H */